Anorexia nervosa From single SNP studies, through biomarkers, to genome-wide association
Marek K. Brandys
ISBN: 978-94-6182-640-4
Printed by: Offpage, Amsterdam
Layout: Marek K. Brandys
Cover design: Marek K. Brandys, based on Effect of Butterfly by Anastasiya Markovich (Picture Labberté K.J.) via Wikimedia Commons
© Marek K. Brandys
Anorexia nervosa From single SNP studies, through biomarkers, to genome-wide association
Anorexia nervosa Van SNP studies via biomarkers naar genoomwijde associatie (met een samenvatting in het Nederlands)
Anorexia nervosa Od badań polimorfizmów pojednynczego nukleotydu, przez biomarkery, po badania asocjacyjne całego genomu (ze streszczeniem w języku polskim)
Proefschrift ter verkrijging van de graad van doctor aan de Universiteit Utrecht op gezag van de rector magnificus, prof.dr. G.J. van der Zwaan, ingevolge het besluit van het college voor promoties in het openbaar te verdedigen op dinsdag 19 januari 2016 des ochtends te 10.30 uur
door
Marek Kajetan Brandys geboren op 27 november 1983 te Kraków, Polska (Polen)
Promotoren: Prof. dr. R.A.H. Adan Prof. dr. A. van Elburg
Copromotoren: Dr. M.J.H. Kas Dr. C. de Kovel
This thesis was partly accomplished with financial support from the Marie Curie Research Training Network INTACT (Individually tailored stepped care for women with eating disorders; reference number: MRTN-CT-2006- 035988)
Table of contents
CHAPTER 1 ...... 6 Introduction Scope and outline of the thesis
CHAPTER 2 ...... 35 Are recently identified genetic variants regulating BMI in the general population associated with anorexia nervosa?
CHAPTER 3 ...... 46 Association study of POMC variants with body composition measures and nutrient choice
CHAPTER 4 ...... 69 Anorexia nervosa and the Val158Met polymorphism of the COMT gene: meta-analysis and new data
CHAPTER 5 ...... 90 A meta-analysis of circulating BDNF concentrations in anorexia nervosa
CHAPTER 6 ...... 129 The Val66Met polymorphism of the BDNF gene in anorexia nervosa: new data and a meta- analysis
CHAPTER 7 ...... 164 No evidence for involvement of CNVs associated with neurodevelopmental disorders in anorexia nervosa
APPENDIX ...... 193 A genome-wide association study of anorexia nervosa
CHAPTER 8 Discussion and conclusions...... 226 Overview of genetic research in anorexia nervosa: the past, the present and the future Concluding remarks
ADDENDUM ...... 264 English summary Nederlandse samenvatting Streszczenie w języku polskim Curriculum Vitae List of publications Acknowledgements
Chapter 1
Chapter 1
Introduction
The main focus of the present thesis is to describe the scientific undertaking of exploring the genetic underpinnings and biomarkers of anorexia nervosa (AN). We begin by introducing the history, intricate phenotypic manifestations, as well as the clinical and epidemiological characteristics of this intriguing disease.
According to DSM-5 AN belongs to the category of feeding and eating disorders (ED), under the code 307.1 (F50.01) for AN restricting type and (F50.02) for the binge-eating/purging type. Other classes in this category include bulimia nervosa (BN; 307.51 (F50.2)), binge eating disorder (BED; 307.51 (F50.8)), other specified feeding or eating disorder (OSFED; 307.59 (F50.8)) and unspecified feeding or eating disorder (307.50 (F50.9)). Diagnostic criteria of AN, according to DSM-5, are listed in a later section.
History It was a British royal physician, Sir William Gull, who in 1873 established the term ‘anorexia’ (derived from Greek ‘an-’, meaning negation, and ‘orexis’, signifying appetite) 1. The first medical descriptions of cases with AN are dated earlier than that, and ascribed to Richard Morton, also a British physician. Looking even further back, there exist historical accounts of people who appear to have suffered from this disorder. In the ancient Hellenistic culture fasting and self-starvation were seen as expressions of religious zealousness. While only a few reports are available from the medieval ages, a larger number of descriptions of the possible cases of AN comes from the times of the Renaissance. Religious ascetics would forge their way to sanity via starvation, self-mutilation and social isolation 2. A
6 Chapter 1 number of historical figures are suspected to have suffered from AN, such as Saint Catherine of Siena, Mary, Queen of Scotts or Elisabeth, Empress of Austria (source: http://divainternational.ch/spip.php?article97). In the modern times, a general interest in AN surged after the death of a famous musician, Karen Carpenter (4 February 1983).
Somatic health risks The most striking feature of the patients suffering from AN is their low body- weight (85% or less than the weight expected). This symptom is accompanied by a refusal to consume sufficient amount of calories – an amount necessary to prevent further emaciation and restoration of the body weight to the normative levels. This continuous undernourishment damages body systems and, in extreme cases, leads to death. Serious medical complications associated with malnutrition in AN include: ° Reduced heart rate and low blood pressure, entailing increased risk of heart failure ° Amenorrhea in post-pubertal females (lack of menstruation) ° Osteoporosis (decreased bone density) ° Loss of muscles ° Dehydration, possibly leading to kidney failure ° Fainting, fatigue, general weakness ° Hair loss, changes of complexion, growth of lanugo – a thin hair layer covering the body Furthermore, health risks associated with the purging behaviors present in the purging subtype of AN include: ° Electrolyte imbalance (caused by dehydration, loss of potassium, chloride and sodium), possibly leading to a heart failure ° Inflammation and possible rupture of esophagus (as a result of vomiting) ° Tooth decay ° Constipation and chronic irregular bowel movements, coming from abuse of laxatives (source:
7 Chapter 1
http://www.nationaleatingdisorders.org/nedaDir/files/documents/h andouts/HlthCons.pdf)
Criteria and symptoms These serious somatic complications are paralleled by the devastation which the disease incurs to the psyche. Suicide is the most frequent cause of death in EDs 3. There is a 57-fold increase in risk of death from suicide among patients with AN, compared to the age-matched cohort 4. Individuals with AN are characterized by the immense fear of weight (fat) gain and disturbed body image. The criteria for diagnosis of AN established in the 5-th edition of The Diagnostic and Statistical Manual of Mental Disorders (DSM-5) are presented below:
A. Restriction of energy intake relative to requirements leading to a significantly low body weight in the context of age, sex, developmental trajectory, and physical health.
B. Intense fear of gaining weight or becoming fat, even though underweight.
C. Disturbance in the way in which one's body weight or shape is experienced, undue influence of body weight or shape on self-evaluation, or denial of the seriousness of the current low body weight.
Symptoms which might suggest AN (warning signs) include (after http://www.allianceforeatingdisorders.com/): • Significant weight loss • Distorted body image • Intense fear/anxiety about gaining weight • Preoccupation with weight, calories, food, etc. • Feelings of guilt after eating • Denial of low weight • High levels of anxiety and/or depression
8 Chapter 1
• Low self-esteem • Self-injury • Withdrawal from friends and activities • Excuses for not eating/denial of hunger • Food rituals • Intense, dramatic mood swings • Pale appearance/yellowish skin-tone • Thin, dull, and dry hair, skin, and nails • Cold intolerance/hypothermia • Fatigue/fainting • Abuse of laxatives, diet pills, or diuretics • Excessive and compulsive exercise
Subtypes and diagnostic cross-over There are two subtypes within the category of AN: the restricting (restrictive) type (ANR) and the bingeing-purging type (ANBP). The main difference is that the individuals with the latter one experience periods of binge eating (consumption of excessive amounts of food coupled with a subjective feeling of loss of control over eating) followed by purging behaviors, which are means to compensate for the calories consumed. Purging can take forms of self- induced vomiting or/and use of laxatives, diuretics or diet pills. Patients exhibiting the restricting type do not purge, but maintain low body weight solely by reduced food intake and increased energy expenditure via exercising and hyperactivity. In the clinical reality, the observed phenotypes are often not clear- cut. This is exemplified by the fact that EDNOS used to be the most frequently assigned diagnosis among EDs (from 40 to 60% of all intakes in ED units, 5 and about 75% of cases of EDs detected among adolescent female population 6), according to DSM-IV criteria. Furthermore, although the category of EDs is relatively stable, moving across the diagnoses within this category is quite common 7. Cases with AN often turn into BN or EDNOS, whereas cases with BN evolve into EDNOS and, much less frequently, into AN
9 Chapter 1
8. Moving between AN subtypes is also frequent – over 7 years, nearly 50% of women diagnosed with AN crossed over from one subtype into another and 34% evolved from AN into BN (with a high chance of relapse into AN) 9. Agras et al. observed that 80% of patients with EDNOS diagnosis had a lifetime history of AN, BN or BED. Additional 10% developed AN, BN or BED during a 4-year follow-up. In this study, EDNOS was a way station between a fully-blown ED and recovery 10. In the most recent, 5-th edition of the DSM the criteria for diagnosis of AN became more inclusive. The amenorrhea criterion has been removed and criterion A became more general. These changes resulted in a better classification of patients with AN and reduction of the vague EDNOS category. A study of 309 patients with ED found that almost all of the 60% of patients with EDNOS according to DSM-IV were re-assigned to the specific diagnoses within the DSM-5 framework 11.
Empirical classifications The outward symptoms ground the division of AN into the ANR and ANBP subtypes. There is a growing body of literature, however, indicating that this classification has limited usefulness for aethiological research and treatment improvement 12. Most of the patients with ANR are going to develop binge- purge behavior at some point, suggesting that these subtypes may in fact represent alternate stages of the same condition, rather than the subtypes 13. Finally, the differences in treatment utilization, relapse and mortality rates are very slight if any 13,14. The observations of the limited usefulness of the clinically-derived subphenotypes drove a number of studies which applied formal statistical procedures to estabilish the experimental classification of eating disorders. Traditional approaches, such as the cluster analysis or the taxometric analysis are currently being replaced by the latent class or latent profile analyses (LPA) 15. In short, LPA employs a maximum likelihood estimation to assign participants to mutually exclusive (unobserved) latent classes. Classes are inferred by the pattern of inter-correlations between indicators (the variables used to infer the classes, e.g. personality traits). It
10 Chapter 1 uses general probability model which allows for inequality of variances in groups and enables determination of the optimal number of classes via formal statistical procedures 15. These analyses were performed in a number of studies on eating disorders (see review 16), and although the particular results will depend on the selected indicators and parameters’ set-up 17, the picture which emerged from these studies is quite consistent 18. In general, the empirical approaches to classification result in the division of the patients with eating disorders into three distinct classes 16,19: 1. the over-regulated and over-controlled class, characterized by constraint and inhibition, 2. the under-regulated and disinhibited subtype, with impulsivity and dysregulated emotional functioning, 3. low psychopathology group, characterized by normative levels of personality functioning and perfectionism. The relevance of these classes has been confirmed in several studies. For instance, Wildes et al. 18 have shown that the empirically derived classification of patients with AN proves useful clinically. The subtypes differed in terms of multiple baseline characteristics, initial response to treatment, readmission rates and outcome at discharge (the undercontrolled patients had worse outcome than the overcontrolled (OR=3.56, p=.01), who, in turn, were worse than the low psychopathology class (OR=11.23, p<.001)). The ANR and ANBP subtypes were of less predictive value. Interestingly, similar classification schemes have been proposed for other psychiatric groups, implying a possible usefulness of classifications based on personality psychopathology across diagnoses 20.
Epidemiology, mortality The epidemiological parameters of AN may differ between the countries and reports. Estimates of the lifetime prevalence of AN in women range from 0.9% to 2.2% 21,22. Incidence among Finnish women from 15th to 19th year of age was 270 per 100,000 person/years 22. Age 15 to 19 is the peak time of onset 23. Point prevalence was estimated at 0.3% 24. These values are based
11 Chapter 1 on a strict definition of AN (according to DSM-IV), and they increase substantially when some of the criteria are relaxed 25. On the whole, although EDs are rare in the general population, they are quite common in young females 26. The incidence of AN has been increasing in the past century until the 1970s and remained stable henceforth 27. AN occurs more often in women (men are affected in only 5% to 10% of cases) 28. AN is notorious for having the highest standardized mortality ratio among psychiatric illnesses (mortality rate being 5 to 10 times higher than in a reference population 29,30. Stratification of patients according to body mass index (height in cm divided by weight in kg squared; BMI) or age of onset shows that SMR is highest in a group of lower BMI and a group of onset later than 17 years of age 31. 20% of individuals with AN who died had committed suicide 32.
Risk factors Although the list of putative risk factors for AN is long 33, the studies examining them are most frequently of a cross-sectional design and hindered by the low frequency of AN in the general population. The focus on the psychosocial factors, rather than on the biological ones, results from the fact that the former are better established in the field and are more easily measurable. A last decade has observed a surge in the number of studies of genetic risk factors. These, however, will be discussed in the other sections. A study with a longitudinal design is preferable when it comes to establishing or verifying putative risk factors. In a prospective study on a birth cohort, Nicholls et al. (2009) tested 22 childhood risk factors proposed in the literature and found that only six were independently associated with development of AN at older age. As expected, female sex turned out to be the most potent risk factor (OR=22.1), followed by history of undereating (OR 2.7), infant feeding problems (OR=2.6), and maternal depressive symptoms (OR 1.8). Conversely, higher self-esteem and higher maternal BMI were found to be protective (OR=0.3 and 0.91, respectively)34. Other studies
12 Chapter 1 add childhood sleeping problems, excessive physical exercise, anxious parenting and perfectionism to this list 35. Another longitudinal study of risk factors in EDs investigated 88 putative factors in a high risk group and found that 7 were independently associated with a chance of developing an ED (critical comments about eating from teacher/coach/siblings and a history of depression had strongest effects on ED risk) 33. Distinguishing a causative risk factor from proxies remains a problematic issue in all studies. There is some evidence supporting the effect of the season of birth on the risk of developing AN, although the effect sizes are small 36. An excess of patients with AN was found among those born in spring (March to June; OR=1.15) 37. The mechanisms underlying the association are not clear. Interestingly, in utero exposure to virus infections (higher incidence of chickenpox and rubella infections) was also related to AN risk (OR=1.6 and 1.5, respectively) 38. Some studies suggested that in utero exposure to male or female steroids may alter the risk of disordered eating in the future 39, but others could not replicate this finding 40.
Comorbidity 87% of patients with EDs 31 had some kind of lifetime psychiatric comorbidity, such as (in order of frequency) depressive disorders, anxiety disorders, suicide attempts, substance abuse disorders, personality disorders (predominantly the borderline personality disorder), obsessive compulsive disorders and others. Suicide attempts were more frequent among ANBP (34%) than in ANR (20%). In about 20% of patients with AN, developmental disorders (autistic spectrum disorder, attention deficit-hyperactivity disorder) are also observed 41. From the range of somatic disorders, which can be comorbid with AN, diabetes mellitus, thyroid disorders and renal calculus are seen most often. Psychiatric and somatic comorbidities are negatively associated with the outcome of AN 30.
13 Chapter 1
Treatment and outcome AN is a disease of a serious social significance. It often runs a chronic course and mainly affects young people 42. Its treatment is expensive 43. Therefore, AN generates substantial direct and indirect costs (for example, the cost of an inpatient treatment for AN in Germany was estimated to be 4647 EUR per case 44). Studies of efficacy and effectiveness of treatment for AN offer only a moderate or low level of evidence 28. Psychotherapeutical approaches which were studied in the context of AN include cognitive-behavioral therapy (CBT), interpersonal therapy, dialectical behavior therapy, psychodynamic therapy, family therapy, adolescent-focused therapy and several others. CBT is the most often recommended modality of treatment (with specific, disease- tailored approaches preferred over non-specific approaches), although the evidence for superiority of any particular approach is far from being conclusive. The main treatment goals include normalization of body weight and eating behaviors and alleviation of psychological problems related to EDs 28. Both outpatient and inpatient settings are used. In cases of extreme emaciation and resistance to treatment, a forced treatment may be used. However, it should be avoided whenever possible. There is little evidence for justification of pharmacotherapy use in AN. Initially promising findings with regards to Olanzapine 45,46 were 47 or were not confirmed in more recent studies 48,49. All these studies were based on small samples and their results are not conclusive. Antidepressive medications have no effect on the course of AN, but they might be used to treat co-morbid depression 28,50. Nonetheless, pharmacological treatment of patients with AN is performed by means of antidepressants (tricyclic and selective serotonin reuptake inhibitors), antipsychotics (typical and atypical), Lithium, naltrexone, antihistamines, clonidine, human growth hormone or cannabis 51. About 57% of patients whose original diagnosis was AN were fully- recovered at a follow-up measurement (the mean duration of the follow-up of 4.8 years) in the Netherlands 52. Another conclusion of this study is that
14 Chapter 1 early detection is associated with a more positive outcome. A study in Germany found that at the 12-year follow-up measurement 27.5% of the patients initially diagnosed with AN had a good outcome, 25.3% an intermediate outcome, 39.6% had a poor outcome, and 7 (7.7%) were deceased 53. Factors associated most strongly with an unfavourable outcome were sexual problems, impulsivity, long duration of inpatient treatment, and long duration of an eating disorder. A review by Steinhausen (2002) adds vomiting, bulimia, and purgative abuse, chronicity of illness, and obsessive- compulsive personality symptoms to the list of unfavourable prognostic features, and notes that other psychiatric disorders at follow-up measurements are very common 42. One of the reasons why therapy of AN is particularly challenging and treatment drop-out rates are high (30-50% 54) is the fact that at least some of the AN symptoms are ego-syntonic. This means that they are in harmony with patients' goals and desires, hence, it is difficult to illicit motivation for treatment. Some of the aspects of AN which might be experienced as rewarding by patients include: • physiological sensations associated with starvation (e.g. stress response leading to endogenous opioid secretion) • gratification from exerting control over body weight and bodily drives (appetite, hunger) • positive feedback from the society or societal groups of reference (e.g. pro-ana groups) • positive feedback from the internalized social mirror (satisfaction of own standards) • hyperactivity might be rewarding (hypothetically, an evolutionary conserved reaction to food scarcity which is supposed to promote foraging) • excessive exercising might be rewarding in several ways • initially rewarding stimuli might lead to adaptation and possible withdrawal effects
15 Chapter 1
Cultural context The modern societies no longer struggle with scarcity of food and the times of famine fade away in the memory of the Western countries. Food has become easily obtainable, both in terms of financial resources and time. Being one of the greatest achievements of the Western civilization, this availability appears to entail increased rates of obesity and, presumably, EDs. There is a stark contrast between ubiquitous food advertisements, which are encouraging overindulging (especially in food that is of low nutritional quality), and the societal pressure to be thin, exercised implicitly or explicitly by our culture. The ambiguous attitude towards food and bodies permeates the modern societies. Of note, some believe that the efforts focused on combating obesity may unintentionally lead to an increase in incidence of EDs. Solid evidence for or against this conviction is lacking. It should be kept in mind that in spite of these conflicting societal pressures and common dissatisfaction with own body (body dissatisfaction in both men and women in western societies is so common that it is considered to be normative 55), only a small fraction of individuals develop an eating disorder. It is a matter of a debate to what extent culture determines EDs. Historical studies and studies on Western and non-Western populations report occurrences of AN without body image concerns or fear of gaining weight (non-fat-phobic AN) 56,57. This means that the sociocultural pressures are neither necessary, nor sufficient for the development of AN 58. Keel & Klump (2003) in their systematic review of the historical and epidemiological data as well as the data coming from studies on non-Western cultures concluded that BN is a more culturally bound condition than AN 59. On the other hand, the incidence of AN was much lower in the Netherlands Antilles than in the Netherlands, but it was found to be similarly common among Netherlands Antilleans living in the Netherlands as among native Dutch 60. The Western idealization of thinness (pressure to be thin) appears to be a risk factor for the development of AN (possibly in interaction with migration- related stress and increased drive to conform in order to counteract alienation) and dieting is a possible triggering factor for the onset of an ED.
16 Chapter 1
Selected candidate molecules for association with AN The alterations of biological functioning in patients with AN are quite dramatic. The difficulty in investigating those lies in determining the difference between premorbid effects (predisposing factors) and effects elicited by starvation and hyperactivity (biomarkers). Thorough discussion of those alterations is beyond the scope of the present thesis (see for example 61,62), but three molecules which are plausible candidates to play a role in AN will be briefly introduced. Brain-derived neurotrophic factor (BDNF) is the most ubiquitous member of the family of neurotrophins. It plays a role in neurodevelopment 63, neural plasticity, connectivity and synaptogenesis 64. It also has been implicated in the regulation of body weight and eating behavior in humans 65 and animals 66. Genome-wide association studies (GWASs) found the BDNF gene locus to be strongly associated with body mass index 67,68. Furthermore, mice with reduced expression of BDNF display increased locomotor activity and aberrant eating behavior leading to obesity 69,70. A hyperphagic phenotype has also been observed in mice with reduced hypothalamic expression of the TrkB – high affinity BDNF receptor 71. Similarly, hyperphagia, obesity and hyperactivity are present in humans who have a functional loss of one copy of the BDNF gene 72. BDNF operates downstream of the melanocortin pathway to regulate energy balance 71. Finally, there is evidence showing BDNF's involvement in reward and stress functioning 73. Catechol-O-methyl transferase (COMT) is an enzyme which degrades catecholamines, such as dopamine and noradrenaline 74. It has been implicated in the pathogenesis of several mental disorders 75. One allele of a functional variant on the COMT gene (rs4680) has been associated with a less stable product and, therefore, lower enzymatic activity 76, which in turn has been hypothesized to lead to higher dopamine availability 77. Rs4680 was studied in mental disorders such as schizophrenia 78, autism 79, depression 80 and eating disorders 81,82. Pro-opiomelanocortin (POMC) is a precursor peptide in the melanocortin system (the melanocortin system is involved in body weight
17 Chapter 1 regulation via effects on appetite and energy expenditure)61. POMC can be cleaved into several important peptides, such as α, β, and γ-MSH and β- Endorphin. Among many other functions, it plays a role in regulation of feeding behaviour 83. BDNF, COMT and POMC and their genetic loci are viable candidates to study in the context of body-weight related phenotypes, especially AN.
The next five sections are based on: Brandys MK, de Kovel CG, Kas MJ, van Elburg AA, Adan RA. Overview of genetic research in anorexia nervosa: The past, the present and the future. Int J Eat Disord 2015. 84
Rationale for gene-association studies Several lines of evidence suggest that there is a substantial genetic component in the aetiology of AN. AN has been observed across many cultures 59. Strong familiar aggregation of AN has been documented (relative risk of 11.3 in first-degree relatives of cases with AN, as compared to the general population 85,86), and the heritability (h2) has been estimated in several twin studies and one adoption study of disordered eating symptoms 87. These estimates range from 0.56 (95% CI, 0.00-0.87) 88 to 0.74 (95% CI: 0.35-0.95) 89, depending on the studied population, definition of AN and applied methodology. The evidence coming from several lines of research demonstrates that the genetic factors are pivotal in the aetiology of AN. No monogenic forms of AN have been found and the data suggest that the genetic underpinning of AN is multifactorial (i.e. multiple genetic variants with small effects, rather than one or a few potent variants, working in concert with environmental factors) 90. Two main types of studies have been employed in a search for those genetic factors. The linkage approach, which investigates co-segregation of the genetic regions with the disease status in large families, has been successful in detecting rare and very potent genetic variants involved in the aetiology of
18 Chapter 1 single-gene disorders (Mendelian), e.g. cystic fibrosis or Huntington’s disease 91,92. However, its usefulness in unravelling common variants of small effects in complex, polygenic diseases or traits remains very limited. The second category is a population-based genetic-association study, which investigates whether frequencies of certain genotypes or alleles are different between cases and controls (significant difference implies association) or if they are correlated with a quantitative trait. This approach focuses on variants with small or medium effects, in a multifactorial model. Within this category, candidate-gene studies (CGSs) look into the single- nucleotide polymorphisms (SNPs) in biologically plausible genes, whereas GWASs test the common SNPs distributed throughout the whole genome.
Candidate gene approach The candidate-gene approach in AN, much like in other psychiatric disorders, turned out to be a primarily futile effort. The scarcity of successful replications can be explained by several reasons, such as genetic differences between the discovery population and the populations in the replication attempts, or by the errors and biases leading to false positive results. Retrospectively, given the complexity and redundancy of biological pathways, and in light of what is now known about the genetic architecture of psychiatric diseases, the hypotheses about which genes could potentially harbor causative mutations had small chances to be proven right. Out of the hundreds of associations indicated by CGSs in the biomedical research only a few were replicated in GWASs 93. This ratio is even less favourable in the field of psychiatry. One study found a lack of enrichment of the association signal in a large genome-wide dataset of cases with schizophrenia and controls after the analysis of 732 autosomal genes indicated in 1374 CGSs (investigation of signal enrichment involves collective testing of a selected group of variants in an independent dataset; it has much greater power, compared to testing of individual variants) 94.
19 Chapter 1
Candidate gene studies in anorexia nervosa Comprehensive reviews of CGSs in AN are available elsewhere 95,96. Although the selection of candidate genes for studies of AN was based on interesting hypotheses 97, and more than 200 gene-association studies were performed in the context of EDs, up to date none of the initially promising findings have been convincingly replicated in the subsequent candidate or genome-wide studies. Meta-analyses, which summarized and weighted the evidence from multiple studies, were also disillusioning 98-101. Also the relatively recent CGS which used the modern standards of design, quality control and statistical significance was negative 102. Still, there are a few findings which await replication attempts, such as rs1473473 of TPH2 103, the 5-HTTLPR polymorphism on SLC6A4 104, rs7180942 in NTRK3 105 and Ala67Thr variant in AGRP 106 (these polymorphisms were not tested in two recent GWASs of AN, because they were not present on the genotyping arrays used in those studies). In parallel to the growing disillusionment about the candidate-gene method, a new approach towards investigation of genetic associations emerged. GWAS technology is relatively recent (first GWAS dates back to 2005 107), but it already has had significant impact on the landscape of biomedical research and resulted in progression of the aetiological knowledge about diseases and traits 108.
Genome-wide association approach GWAS is a hypothesis-free approach. It uses microarray platforms to examine the genotypic data from a large number of SNPs (from hundreds of thousands up to millions), which cover most of the human common SNP variation (a SNP is considered common if the frequency of its minor allele is larger than 1%). This coverage is increased via imputation - a procedure which uses statistical algorithms to infer the genotypes of the ungenotyped SNPs by employing the reference data coming from e.g. HapMap or 1000 Genomes Project populations. Genome-wide data also allows for
20 Chapter 1 investigation of copy number variants (CNVs; deleted or duplicated stretches of the genome).
Below is a list of the main goals of GWASs: • Furthering the understanding of the biological mechanisms of the disease, by finding the genes and pathways involved in the aetiology. This is the foremost goal of GWASs. • Learning about the genetic architecture. This includes the expected range of effect sizes, allelic frequencies of the associated variants, underlying genetic models (additive, dominant, recessive, overdominant, multiplicative) and the possibility of gene x environment and gene x gene interactions. • Understanding of the genetic overlap between diseases and traits. This has a potential of enhancing the nosological system and treatment. • Genetic screening to identify populations at risk (risk prediction) or individual genotyping of a patient to inform diagnosis and treatment (personalized medicine). As exciting as these prospects are, they are distant goals, and in view of a highly polygenic nature of psychiatric diseases, they are unlikely to be achievable in the near future 109.
What needs to be remembered when interpreting a GWAS is that its results inform about association but do not determine causality, and that a statistical strength of association at a given locus should not be confused with its biological relevance (the most significant finding in GWASs might not be the most informative).
Scope and outline of the thesis
The overarching theme of this thesis is the effort to shed light on the genetic background of AN. That undertaking predominantly involves searching for
21 Chapter 1 the genetic associations via the candidate-gene and genome-wide studies. Thus, we aimed to find the genetic variants which change the risk of developing AN, and to increase the understanding of the mechanisms of the disease. Beyond investigating the genetic associations, in one chapter we also take interest in studies examining the serum levels of the BDNF neurotrophin in patients with AN. BDNF is a product of the BDNF gene which was also investigated in this thesis.
We begin by studying a possible genetic relation of AN and obesity by testing several genetic variants associated with the latter in a sample of patients with AN110. This work adds to the discussion about the genetic nature of AN and its hypothesized relation to the opposite extreme of the weight spectrum - obesity (chapter 2). The next chapter (3) presents the only study which does not directly involve patients with AN. In that publication we describe the search for association of variants from the POMC locus with detailed measures of body composition and nutrient choice in the general population111. POMC molecule, due to its effects on the appetite regulation, was a plausible candidate for having a role in AN diathesis. We also explore the alterations of biological functioning in AN, in the context of possible candidates for genetic associations. Chapter 4 is a meta- analysis of several studies which compared the serum BDNF levels in patients with AN and healthy controls112. A discussion of whether the observed effects are state or trait dependent is included. Thereafter, in chapter 5, we apply the meta-analytical methodology to a promising candidate for genetic association with AN, the Val66Met polymorphism on the BDNF gene101. Meta-analyses combine evidence from multiple studies on a single subject, which (together with our own, novel data) allows us to draw stronger conclusions than any of these studies alone. The further investigations (chapter 6) consider a genetic polymorphism long thought to play a role in several mental disorders, including ED. Val158Met polymorphism of the COMT gene was tested in the
22 Chapter 1 new data and the results were merged in a meta-analytical framework with the results of previous studies on this subject100. Chapter 7 presents the work which used the genome-wide data of patients with AN and controls to test for association of AN with selected structural variants (rather than SNPs). The nature of the available data was not sufficient to render the results fully credible; hence this study remains a preliminary investigation. The appendix describes a large collaborative study which uses thousands of DNA samples from all over the world and analyses them in a hypothesis-free, genome-wide approach113. This remains the largest genetic study in AN up to date. The final publication included in this dissertation (chapter 8) is an opinion paper which reviews and discusses the past approaches to the studies of gene-association in AN, shows how they evolved over the years (a process well reflected in the present thesis), and tries to outline the directions for future research.
References 1. Gull WW. Anorexia nervosa (apepsia hysterica, anorexia hysterica). 1868. Obes Res 1997;5:498-502. 2. Pearce JMS. Richard morton: Origins of anorexia nervosa. Eur Neurol 2004;52:191-192. 3. Sullivan P. Mortality in anorexia nervosa. Am J Psychiatry 1995;152:1073- 1074. 4. Keel PK, Dorer DJ, Eddy KT, Franko D, Charatan DL, Herzog DB. Predictors of mortality in eating disorders. Arch Gen Psychiatry 2003;60:179-183. 5. Thomas JJ, Vartanian LR, Brownell KD. The relationship between eating disorder not otherwise specified (EDNOS) and officially recognized eating disorders: Meta-analysis and implications for DSM. Psychol Bull 2009;135:407-433. 6. Machado PPP, Machado BC, Gonçalves S, Hoek HW. The prevalence of eating disorders not otherwise specified. Int J Eat Disord 2007;40:212-217.
23 Chapter 1
7. Milos G, Spindler A, Schnyder U, Fairburn CG. Instability of eating disorder diagnoses: Prospective study. The British Journal of Psychiatry 2005;187:573- 578. 8. Tozzi F, Thornton LM, Klump KL, Fichter MM, Halmi KA, Kaplan AS, et al. Symptom fluctuation in eating disorders: Correlates of diagnostic crossover. Am J Psychiatry 2005;162:732-740. 9. Eddy KT, Dorer DJ, Franko DL, Tahilani K, Thompson-Brenner H, Herzog DB. Diagnostic crossover in anorexia nervosa and bulimia nervosa: Implications for DSM-V. Am J Psychiatry 2008;165:245-250. 10. Agras WS, Crow S, Mitchell JE, Halmi KA, Bryson S. A 4-year prospective study of eating disorder NOS compared with full eating disorder syndromes. Int J Eat Disord 2009;42:565-570. 11. Fisher MM, Gonzalez M, Malizio J. Eating disorders in adolescents: How does the DSM-5 change the diagnosis? Journal of Adolescent Health 2014;54:S9. 12. Peat C, Mitchell JE, Hoek H, Wonderlich S. Validity and utility of subtyping anorexia nervosa. Int J Eat Disord 2009;9999:NA. 13. Eddy KT, Keel PK, Dorer DJ, Delinsky SS, Franko DL, Herzog DB. Longitudinal comparison of anorexia nervosa subtypes. Int J Eat Disord 2002;31:191-201. 14. Keel PK, Dorer DJ, Eddy KT, Delinsky SS, Franko DL, Blais MA, et al. Predictors of treatment utilization among women with anorexia and bulimia nervosa. Am J Psychiatry 2002;159:140-142. 15. Magidson J, Vermunt JK. Latent class models for clustering: A comparison with K-means. Canadian Journal of Marketing Research 2002;20:36-36-43. 16. Wonderlich SA, Joiner TE, Keel PK, Williamson DA, Crosby RD. Eating disorder diagnoses: Empirical approaches to classification. Am Psychol 2007;62:167-180. 17. Swanson SA, Lindenberg K, Bauer S, Crosby RD. A monte carlo investigation of factors influencing latent class analysis: An application to eating disorder research. Int J Eat Disord 2011:n/a-n/a.
24 Chapter 1
18. Wildes JE, Marcus MD, Crosby RD, Ringham RM, Dapelo MM, Gaskill JA, et al. The clinical utility of personality subtypes in patients with anorexia nervosa. J Consult Clin Psychol 2011;79:665-674. 19. Steiger H, Richardson J, Schmitz N, Israel M, Bruce KR, Gauvin L. Trait- defined eating-disorder subtypes and history of childhood abuse. Int J Eat Disord 2010;43:428-432. 20. Miller MW, Kaloupek DG, Dillon AL, Keane TM. Externalizing and internalizing subtypes of combat-related PTSD: A replication and extension using the PSY-5 scales. J Abnorm Psychol 2004;113:636-645. 21. Hudson JI, Hiripi E, Pope Jr. HG, Kessler RC. The prevalence and correlates of eating disorders in the national comorbidity survey replication. Biological Psychiatry 2007;61:348-358. 22. Keski-Rahkonen A, Hoek HW, Susser ES, Linna MS, Sihvola E, Raevuori A, et al. Epidemiology and course of anorexia nervosa in the community. Am J Psychiatry 2007;164:1259-1265. 23. Lucas A, Beard C, O'Fallon W, Kurland L. 50-year trends in the incidence of anorexia nervosa in rochester, minn.: A population-based study. Am J Psychiatry 1991;148:917-922. 24. Hoek HW, van Hoeken D. Review of the prevalence and incidence of eating disorders. International Journal of Eating Disorders 2003;34:383-396. 25. Isomaa R, Isomaa A, Marttunen M, Kaltiala-Heino R, Björkqvist K. The prevalence, incidence and development of eating disorders in finnish adolescents - a two-step 3-year follow-up study. European Eating Disorders Review 2009;17:199-207. 26. Smink FR, van Hoeken D, Hoek HW. Epidemiology of eating disorders: Incidence, prevalence and mortality rates. Curr Psychiatry Rep 2012;14:406- 414. 27. Currin L, Schmidt U, Treasure J, Jick H. Time trends in eating disorder incidence. The British Journal of Psychiatry 2005;186:132-135. 28. Herpertz S, Hagenah U, Vocks S, von Wietersheim J, Cuntz U, Zeeck A. The diagnosis and treatment of eating disorders. Dtsch Arztebl Int 2011;108:678-685.
25 Chapter 1
29. Birmingham CL, Su J, Hlynsky JA, Goldner EM, Gao M. The mortality rate from anorexia nervosa. Int J Eat Disord 2005;38:143-146. 30. Papadopoulos FC, Ekbom A, Brandt L, Ekselius L. Excess mortality, causes of death and prognostic factors in anorexia nervosa. The British Journal of Psychiatry 2009;194:10-17. 31. Rosling AM, Sparén P, Norring C, von Knorring A. Mortality of eating disorders: A follow-up study of treatment in a specialist unit 1974?2000. Int J Eat Disord 2011;44:304-310. 32. Arcelus J, Mitchell AJ, Wales J, Nielsen S. Mortality rates in patients with anorexia nervosa and other eating disorders: A meta-analysis of 36 studies. Arch Gen Psychiatry 2011;68:724-731. 33. Jacobi C, Fittig E, Bryson SW, Wilfley D, Kraemer HC, Taylor CB. Who is really at risk? identifying risk factors for subthreshold and full syndrome eating disorders in a high-risk sample. Psychol Med 2011;41:1939. 34. Nicholls DE, Viner RM. Childhood risk factors for lifetime anorexia nervosa by age 30 years in a national birth cohort. J Am Acad Child Adolesc Psychiatry 2009;48:791-799. 35. Pike KM, Hilbert A, Wilfley DE, Fairburn CG, Dohm FA, Walsh BT, et al. Toward an understanding of risk factors for anorexia nervosa: A case-control study. Psychol Med 2008;38:1443-1453. 36. Winje E, Willoughby K, Lask B. Season of birth bias in eating disorders-- fact or fiction? Int J Eat Disord 2008;41:479-490. 37. Disanto G, Handel AE, Para AE, Ramagopalan SV, Handunnetthi L. Season of birth and anorexia nervosa. The British Journal of Psychiatry 2011. 38. Favaro A, Tenconi E, Ceschin L, Zanetti T, Bosello R, Santonastaso P. In utero exposure to virus infections and the risk of developing anorexia nervosa. Psychol Med 2011:1-7. 39. Culbert KM, Breedlove SM, Burt SA, Klump KL. Prenatal hormone exposure and risk for eating disorders: A comparison of opposite-sex and same-sex twins. Arch Gen Psychiatry 2008;65:329-336.
26 Chapter 1
40. Baker JH, Lichtenstein P, Kendler KS. Intrauterine testosterone exposure and risk for disordered eating. The British Journal of Psychiatry 2009;194:375-376. 41. Wentz E, Lacey JH, Waller G, Rastam M, Turk J, Gillberg C. Childhood onset neuropsychiatric disorders in adult eating disorder patients. A pilot study. Eur Child Adolesc Psychiatry 2005;14:431-437. 42. Steinhausen H. The outcome of anorexia nervosa in the 20th century. Am J Psychiatry 2002;159:1284-1293. 43. Striegel-Moore RH, Leslie D, Petrill SA, Garvin V, Rosenheck RA. One-year use and cost of inpatient and outpatient services among female and male patients with an eating disorder: Evidence from a national database of health insurance claims. Int J Eat Disord 2000;27:381-389. 44. Haas L, Stargardt T, Schreyoegg J, Schlösser R, Danzer G, Klapp BF. Inpatient costs and predictors of costs in the psychosomatic treatment of anorexia nervosa. Int J Eat Disord 2011:n/a-n/a. 45. Brambilla F, Garcia CS, Fassino S, Daga GA, Favaro A, Santonastaso P, et al. Olanzapine therapy in anorexia nervosa: Psychobiological effects. Int Clin Psychopharmacol 2007;22:197-204. 46. Leggero C, Masi G, Brunori E, Calderoni S, Carissimo R, Maestro S, et al. Low-dose olanzapine monotherapy in girls with anorexia nervosa, restricting subtype: Focus on hyperactivity. J Child Adolesc Psychopharmacol 2010;20:127-133. 47. Attia E, Kaplan AS, Walsh BT, Gershkovich M, Yilmaz Z, Musante D, et al. Olanzapine versus placebo for out-patients with anorexia nervosa. Psychol Med 2011:1-6. 48. Kafantaris V, Leigh E, Hertz S, Berest A, Schebendach J, Sterling WM, et al. A placebo-controlled pilot study of adjunctive olanzapine for adolescents with anorexia nervosa. J Child Adolesc Psychopharmacol 2011;21:207-212. 49. Norris ML, Spettigue W, Buchholz A, Henderson KA, Gomez R, Maras D, et al. Olanzapine use for the adjunctive treatment of adolescents with anorexia nervosa. J Child Adolesc Psychopharmacol 2011.
27 Chapter 1
50. Suarez-Pinilla P, Pena-Perez C, Arbaizar-Barrenechea B, Crespo-Facorro B, Del Barrio JA, Treasure J, et al. Inpatient treatment for anorexia nervosa: A systematic review of randomized controlled trials. J Psychiatr Pract 2015;21:49-59. 51. Aigner M, Treasure J, Kaye W, Kasper S. World federation of societies of biological psychiatry (WFSBP) guidelines for the pharmacological treatment of eating disorders. World Journal of Biological Psychiatry 2011;12:400-443. 52. van Son GE, van Hoeken D, van Furth EF, Donker GA, Hoek HW. Course and outcome of eating disorders in a primary care-based cohort. Int J Eat Disord 2009;9999:NA. 53. Fichter MM, Quadflieg N, Hedlund S. Twelve-year course and outcome predictors of anorexia nervosa. Int J Eat Disord 2006;39:87-100. 54. Guarda AS. Treatment of anorexia nervosa: Insights and obstacles. Physiology & Behavior, 2008;94:113-120. 55. Tantleff-Dunn S, Barnes RD, Larose JG. It's not just a "woman thing:" The current state of normative discontent. Eat Disord 2011;19:392-402. 56. Santonastaso P, Bosello R, Schiavone P, Tenconi E, Degortes D, Favaro A. Typical and atypical restrictive anorexia nervosa: Weight history, body image, psychiatric symptoms, and response to outpatient treatment. Int J Eat Disord 2009;42:464-470. 57. Becker AE, Thomas JJ, Pike KM. Should non-fat-phobic anorexia nervosa be included in DSM-V? Int J Eat Disord 2009;42:620-635. 58. Schmidt U. Epidemiology and aetiology of eating disorders. Psychiatry, 2005;4:5-9. 59. Keel PK, Klump KL. Are eating disorders culture-bound syndromes? implications for conceptualizing their etiology. Psychol Bull 2003;129:747- 769. 60. van Hoeken D, Veling W, Smink FRE, Hoek HW. The incidence of anorexia nervosa in netherlands antilles immigrants in the netherlands. Eur Eat Disord Rev 2010;18:399-403. 61. Miller K. Endocrine dysregulation in anorexia nervosa update. The Journal of Clinical Endocrinology & Metabolism 2011;96:2939-2949.
28 Chapter 1
62. Adan RAH, Hillebrand JJG, Danner UN, Cano SC, Kas MJH, Verhagen LAW. Neurobiology driving hyperactivity in activity-based anorexia. Curr Top Behav Neurosci 2011;6:229-250. 63. Hyman C, Hofer M, Barde Y, Juhasz M, Yancopoulos GD, Squinto SP, et al. BDNF is a neurotrophic factor for dopaminergic neurons of the substantia nigra. Nature 1991;350:230-232. 64. Greenberg ME, Xu B, Lu B, Hempstead BL. New insights in the biology of BDNF synthesis and release: Implications in CNS function. J Neurosci 2009;29:12764-12767. 65. Lommatzsch M, Zingler D, Schuhbaeck K, Schloetcke K, Zingler C, Schuff- Werner P, et al. The impact of age, weight and gender on BDNF levels in human platelets and plasma. Neurobiol Aging 2005;26:115-123. 66. King BM. The rise, fall, and resurrection of the ventromedial hypothalamus in the regulation of feeding behavior and body weight. Physiol Behav 2006;87:221-244. 67. Willer CJ, Speliotes EK, Loos RJ, Li S, Lindgren CM, Heid IM, et al. Six new loci associated with body mass index highlight a neuronal influence on body weight regulation. Nat Genet 2009;41:25-34. 68. Locke AE, Kahali B, Berndt SI, Justice AE, Pers TH, Day FR, et al. Genetic studies of body mass index yield new insights for obesity biology. Nature 2015;518:197-206. 69. Kernie SG, Liebl DJ, Parada LF. BDNF regulates eating behavior and locomotor activity in mice. EMBO J 2000;19:1290-1300. 70. Rios M, Fan G, Fekete C, Kelly J, Bates B, Kuehn R, et al. Conditional deletion of brain-derived neurotrophic factor in the postnatal brain leads to obesity and hyperactivity. Mol Endocrinol 2001;15:1748-1757. 71. Xu B, Goulding EH, Zang K, Cepoi D, Cone RD, Jones KR, et al. Brain- derived neurotrophic factor regulates energy balance downstream of melanocortin-4 receptor. Nat Neurosci 2003;6:736-742. 72. Gray J, Yeo GSH, Cox JJ, Morton J, Adlam AR, Keogh JM, et al. Hyperphagia, severe obesity, impaired cognitive function, and hyperactivity
29 Chapter 1 associated with functional loss of one copy of the brain-derived neurotrophic factor (BDNF) gene. Diabetes 2006;55:3366-3371. 73. Narita M, Aoki K, Takagi M, Yajima Y, Suzuki T. Implication of brain- derived neurotrophic factor in the release of dopamine and dopamine- related behaviors induced by methamphetamine. Neuroscience 2003;119:767-775. 74. Chen J, Lipska BK, Halim N, Ma QD, Matsumoto M, Melhem S, et al. Functional analysis of genetic variation in catechol-O-methyltransferase (COMT): Effects on mRNA, protein, and enzyme activity in postmortem human brain. Am J Hum Genet 2004;75:807-821. 75. Mier D, Kirsch P, Meyer-Lindenberg A. Neural substrates of pleiotropic action of genetic variation in COMT: A meta-analysis. Mol Psychiatry 2010;15:918-927. 76. Lachman HM, Papolos DF, Saito T, Yu YM, Szumlanski CL, Weinshilboum RM. Human catechol-O-methyltransferase pharmacogenetics: Description of a functional polymorphism and its potential application to neuropsychiatric disorders. Pharmacogenetics 1996;6:243-250. 77. Shield AJ, Thomae BA, Eckloff BW, Wieben ED, Weinshilboum RM. Human catechol O-methyltransferase genetic variation: Gene resequencing and functional characterization of variant allozymes. Mol Psychiatry 2004;9:151-160. 78. Costas J, Sanjuán J, Ramos-Ríos R, Paz E, Agra S, Ivorra JL, et al. Heterozygosity at catechol-O-methyltransferase Val158Met and schizophrenia: New data and meta-analysis. J Psychiatr Res 2011;45:7-14. 79. Stergiakouli E, Thapar A. Fitting the pieces together: Current research on the genetic basis of attention-deficit/hyperactivity disorder (ADHD). Neuropsychiatr Dis Treat 2010;6:551-560. 80. Åberg E, Fandiño-Losada A, Sjöholm LK, Forsell Y, Lavebratt C. The functional Val158Met polymorphism in catechol-O-methyltransferase (COMT) is associated with depression and motivation in men from a swedish population-based study. J Affect Disord 2011;129:158-166.
30 Chapter 1
81. Mikolajczyk E, Smiarowska M, Grzywacz A, Samochowiec J. Association of eating disorders with catechol- O -methyltransferase gene functional polymorphism. Neuropsychobiology 2006;54:82-86. 82. Frieling H, Romer KD, Wilhelm J, Hillemacher T, Kornhuber J, de Zwaan M, et al. Association of catecholamine-O-methyltransferase and 5-HTTLPR genotype with eating disorder-related behavior and attitudes in females with eating disorders. Psychiatr Genet 2006;16:205-208. 83. Millington G. The role of proopiomelanocortin (POMC) neurones in feeding behaviour. Nutr Metab (Lond) 2007;4:18. 84. Brandys MK, de Kovel CG, Kas MJ, van Elburg AA, Adan RA. Overview of genetic research in anorexia nervosa: The past, the present and the future. Int J Eat Disord 2015. 85. Lilenfeld LR, Kaye WH, Greeno CG, Merikangas KR, Plotnicov K, Pollice C, et al. A controlled family study of anorexia nervosa and bulimia nervosa: Psychiatric disorders in first-degree relatives and effects of proband comorbidity. Arch Gen Psychiatry 1998;55:603-610. 86. Strober M, Freeman R, Lampert C, Diamond J, Kaye W. Controlled family study of anorexia nervosa and bulimia nervosa: Evidence of shared liability and transmission of partial syndromes. Am J Psychiatry 2000;157:393-401. 87. Klump KL, Suisman JL, Burt SA, McGue M, Iacono WG. Genetic and environmental influences on disordered eating: An adoption study. J Abnorm Psychol 2009;118:797. 88. Bulik CM, Sullivan PF, Tozzi F, Furberg H, Lichtenstein P, Pedersen NL. Prevalence, heritability, and prospective risk factors for anorexia nervosa. Arch Gen Psychiatry 2006;63:305-312. 89. Klump KL, Miller K, Keel P, McGue M, Iacono W. Genetic and environmental influences on anorexia nervosa syndromes in a population–based twin sample. Psychol Med 2001;31:737-740. 90. Steinhausen H, Jakobsen H, Helenius D, Munk-Jorgensen P, Strober M. A nationwide study of the family aggregation and risk factors in anorexia nervosa over three generations. Int J Eat Disord 2014;48:1-8.
31 Chapter 1
91. Knowlton RG, Cohen-Haguenauer O, Van Cong N, Frezal J, Brown VA, Barker D, et al. A polymorphic DNA marker linked to cystic fibrosis is located on chromosome 7. Nature 1985;318:380-382. 92. Gusella JF, Wexler NS, Conneally PM, Naylor SL, Anderson MA, Tanzi RE, et al. A polymorphic DNA marker genetically linked to huntington's disease. Nature 1983;306:234-238. 93. Siontis KC, Patsopoulos NA, Ioannidis JP. Replication of past candidate loci for common diseases and phenotypes in 100 genome-wide association studies. European Journal of Human Genetics 2010;18:832-837. 94. Collins AL, Kim Y, Sklar P, International Schizophrenia Consortium, O'Donovan MC, Sullivan PF. Hypothesis-driven candidate genes for schizophrenia compared to genome-wide association results. Psychol Med 2012;42:607-616. 95. Clarke TK, Weiss ARD, Berrettini WH. The genetics of anorexia nervosa. Clin Pharmacol Ther 2011;91:181-188. 96. Rask-Andersen M, Olszewski PK, Levine AS, Schioth HB. Molecular mechanisms underlying anorexia nervosa: Focus on human gene association studies and systems controlling food intake. Brain Res Rev 2010;62:147-164. 97. Hinney A, Volckmar A. Genetics of eating disorders. Curr Psychiatry Rep 2013;15:1-9. 98. Slof-Op't Landt MC, Furth EF, Meulenbelt I, Bartels M, Hottenga JJ, Slagboom PE, et al. Association study of the estrogen receptor I gene (ESR1) in anorexia nervosa and eating disorders: No replication found. Int J Eat Disord 2013;47:211-214. 99. Zhang C, Chen J, Jia X, Yu S, Jiang W, Zhang R, et al. Estrogen receptor 1 gene rs2295193 polymorphism and anorexia nervosa: New data and meta†analysis. Asia†Pacific Psychiatry 2013;5:331-335. 100. Brandys MK, Slof-Op't Landt MC, van Elburg AA, Ophoff R, Verduijn W, Meulenbelt I, et al. Anorexia nervosa and the Val158Met polymorphism of the COMT gene: Meta-analysis and new data. Psychiatr Genet 2012;22:130- 136.
32 Chapter 1
101. Brandys MK, Kas MJ, van Elburg AA, Ophoff R, Slof-Op't Landt MC, Middeldorp CM, et al. The Val66Met polymorphism of the BDNF gene in anorexia nervosa: New data and a meta-analysis. The World Journal of Biological Psychiatry 2013;14:441-451. 102. Pinheiro AP, Bulik CM, Thornton LM, Sullivan PF, Root TL, Bloss CS, et al. Association study of 182 candidate genes in anorexia nervosa. Am J Med Genet B Neuropsychiatr Genet 2010;153B:1070-1080. 103. Slof-O't Landt M, Meulenbelt I, Bartels M, Suchiman E, Middeldorp C, Houwing†Duistermaat J, et al. Association study in eating disorders: TPH2 associates with anorexia nervosa and self-â induced vomiting. Genes, Brain and Behavior 2011;10:236-243. 104. Calati R, De Ronchi D, Bellini M, Serretti A. The 5-HTTLPR polymorphism and eating disorders: A meta- analysis. Int J Eat Disord 2011;44:191-199. 105. Mercader JM, Saus E, Aguera Z, Bayes M, Boni C, Carreras A, et al. Association of NTRK3 and its interaction with NGF suggest an altered cross- regulation of the neurotrophin signaling pathway in eating disorders. Hum Mol Genet 2008;17:1234-1244. 106. Vink T, Hinney A, Van Elburg A, van Goozen SH, Sandkuijl L, Sinke R, et al. Association between an agouti-related protein gene polymorphism and anorexia nervosa. Mol Psychiatry 2001;6:325-328. 107. Klein RJ, Zeiss C, Chew EY, Tsai JY, Sackler RS, Haynes C, et al. Complement factor H polymorphism in age-related macular degeneration. Science 2005;308:385-389. 108. Kim Y, Zerwas S, Trace SE, Sullivan PF. Schizophrenia genetics: Where next? Schizophr Bull 2011;37:456-463. 109. Wray NR, Yang J, Hayes BJ, Price AL, Goddard ME, Visscher PM. Pitfalls of predicting complex traits from SNPs. Nature Reviews Genetics 2013;14:507-515. 110. Brandys MK, van Elburg AA, Loos RJF, Bauer F, Hendriks J, van der Schouw YT, et al. Are recently identified genetic variants regulating BMI in the general population associated with anorexia nervosa? American Journal of Medical Genetics Part B: Neuropsychiatric Genetics 2010;153B:695-699.
33 Chapter 1
111. Ternouth A, Brandys MK, van der Schouw YT, Hendriks J, Jansson J, Collier D, et al. Association study of POMC variants with body composition measures and nutrient choice. Eur J Pharmacol 2011. 112. Brandys MK, Kas MJH, van Elburg AA, Campbell IC, Adan RAH. A meta- analysis of circulating BDNF concentrations in anorexia nervosa. World J Biol Psychiatry 2011. 113. Boraska V, Franklin C, Floyd J, Thornton L, Huckins L, Southam L, et al. A genome-wide association study of anorexia nervosa. Mol Psychiatry 2014;19:1085-1094.
34 Chapter 2
Chapter 2
Are recently identified genetic variants regulating BMI in the general population associated with anorexia nervosa?
Marek K. Brandys Annemarie A. van Elburg Ruth J.F. Loos Florianne Bauer Judith Hendriks Yvonne T. van der Schouw Roger A.H. Adan
American Journal of Medical Genetics Part B: Neuropsychiatric Genetics 2010; 153B(2): 695-699.
35 Chapter 2
Abstract
The influence of body mass index (BMI) on susceptibility to anorexia nervosa (AN) is not clear. Recently published genome-wide association (GWA) studies of the general population identified several variants influencing BMI. We genotyped these variants in an AN sample to test for association and to investigate a combined effect of BMI-increasing alleles (as determined in the original GWA studies) on the risk of developing the disease. Individual single nucleotide polymorphisms (SNPs) were tested for association with AN in a sample of 267 AN patients and 1636 population controls. A logistic regression for the combined effect of BMI-increasing alleles included 225 cases and 1351 controls. We found no significant association between individual SNPs and AN. The analysis of a combined effect of BMI-increasing alleles showed absence of association with the investigated condition. The percentages of BMI-increasing alleles were equal between cases and controls. This study found no evidence that genetic variants regulating BMI in the general population are significantly associated with susceptibility to AN.
36 Chapter 2
Introduction
Two extensive population-based genome-wide association studies (GWA) of BMI and obesity have been published recently 1,2. Both studies revealed associations of new loci and confirmed already known roles of FTO and MC4R 3,4. All together, these loci harbor 10 genes, most of them predominantly expressed in the central nervous system (Table I). Although little is known about the function of the genes linked to the newly identified genetic variants for body mass index (BMI) and common obesity, preliminary evidence suggests that they might affect BMI via involvement in the neuronal regulation of food intake 1,2. Additionally, it has been proposed that eating disorders and obesity may be considered on the same continuum of psychopathology (as opposed to discrete models; 5). We, therefore, hypothesized that variants affecting BMI and obesity may potentially alter the risk of developing an eating disorder such as anorexia nervosa (AN). It is under debate whether high or low BMI before the onset of the disease has an impact upon susceptibility to AN 6,7. AN patients with lower premorbid BMI (self-reported) tend to present with lower BMI at first referral for the treatment 8. A recent study found a correlation between premorbid BMI and BMI at discharge from the treatment and at follow-up in AN 6,8. Patients with lower BMI before the onset of the disease and at admission had poorer general indices of functioning 6. Elevated BMI could either protect against the disease – since AN is a disorder of low body weight – or increase the risk, via a tendency to a general eating pathology such as e.g. restrained eating, excessive dieting or a persistent desire to lose weight. In the present study we aimed to test whether the genetic variants increasing BMI in the general population escalate the susceptibility to AN or diminish it (or have no effect upon it).
37 Chapter 2
TABLE I. Characteristics of investigated SNPs and analysis of association with AN (allelic test, 1df)
Effect allele Effect BMI Effect allele Effect sizes detectable at 80% power OR for Nearby freq. reported allele HWE in SNP incr. freq. in the assoc. P-value, allelic test gene in referred freq. in cont. OR for hetero- OR for homo- allele 1,2 controls (95%CI) GWAS cases zygotes zygotes .96 rs1121980* FTO A 41% 41.3% 42.1% .10 1.3 1.7 .73 (.80-1.16) .97 rs17700633 MC4R A 32% 29.5% 30.1% .99 1.3 1.7 .78 (.79-1.18) .99 rs17782313 MC4R C 21% 26.0% 26.1% 1.00 1.4 1.7 .96 (.80-1.22) TMEM1 1.03 rs6548238 C 84% 83.3% 83.7% .86 1.5 2.3 .81 8 (.80-1.32) GNPDA 1.00 rs10938397 G 45% 42.2% 42.0% .98 1.3 1.7 .94 2 (.83-1.21) 1.02 rs7498665 SH2B1 G 41% 38.5% 37.8% .86 1.3 1.7 .79 (.84-1.23) .95 rs368794* KCTD15 T 68% 68.5% 67.5% 1.00 1.3 1.8 .67 (.78-1.17) 1.06 rs10838738 MTCH2 G 34% 34.0% 32.5% .45 1.3 1.7 .51 (.87-1.29) .90 rs2568958* NEGR1 A 62% 60.2% 57.9% .30 1.3 1.8 .32 (.75-1.09) 1.13 rs1488830* BDNF T 79% 75.9% 78.0% .69 1.5 2.1 .26 (.91-1.40) .95 rs925946 BDNF T 30% 28.1% 29.2% .57 1.3 1.8 .62 (.77-1.16) 1.01 rs7647305 ETV5 C 80% 79.9% 80.1% .93 1.5 2.2 .93 (.80-1.27) SNP, Single Nucleotide Polymorphism; BMI, body mass index; Freq., frequency; Cont., controls; OR, odds ratio; CI, confidence intervals; HWE, Hardy-Weinberg equilibrium; χ2 test with 1df for HWE; assumptions for power calculation: allelic test (1df), α =.05, prevalence =.02. * SNPs in LD with SNPs identified in GWA studies of BMI (proxies)
38 Chapter 2
Methods and materials
A total of 13 BMI-associated single nucleotide polymorphisms (SNP), selected from the recent GWA studies of BMI 1-4, were genotyped in 267 AN patients and 1636 control individuals. Nine SNPs were the same as those identified in the GWA studies and four SNPs were in perfect or high linkage disequilibrium (LD; r2 > 0.84) with the SNPs of interest (Table I). The patients’ group consisted of female AN cases with ascertained Dutch descent (patients are asked whether all of their grandparents were of Dutch origin). There were 182 AN restrictive type and 99 AN purging type cases. Subjects were recruited for the study after referral to Eating Disorders treatment center (in- and outpatients, at various stages of the disease). Diagnosis was established by experienced clinicians according to DSM-IV criteria, with use of a semi-structured interview (Eating Disorder Examination; 9). Cases in which AN was not the primary diagnosis or with physical illnesses such as diabetes mellitus were excluded. The control group consisted of a random sample of Dutch female participants in the Utrecht contribution to the European Prospective Investigation into Cancer and Nutrition, also known as Prospect-EPIC 10. Lack of selection criteria is balanced by a relatively large size of this random population sample. Fourteen (5%) out of 281 cases and twenty (1%) out of 1656 controls were excluded because of more than two missing genotypes. In the remaining 267 cases and 1636 controls the mean age (SD) was 22.4 (4.3) years and 49.0 (6.0) years, and the mean BMI was 16.4 (2.1) kg/m2 and 25.9 (4.0) kg/m2, respectively. Genotyping call rate among successfully genotyped individuals was 98.4 %. Apart from SNP rs2844479, which was excluded from further analysis because of significant difference in missing calls between cases and controls, all SNPs passed the quality control requirements (more than 95% successful calls per SNP, Hardy-Weinberg Equilibrium test (χ2; 1 degree of freedom (df))
39 Chapter 2 p > .01, minor allele frequency > .05, difference in missing calls between groups at p>.01). To further ensure quality, blind duplicates were included on plates (100% concordance of duplicates, excluding missed calls). Genotyping was performed on a commercial platform (KBiosciences; Hertsfordshire, U.K.). Statistical analyses were conducted with PLINK 11, UNPHASED 12 and SPSS 15.0 (SPSS, Chicago, Illinois). In the first step of analysis we performed case-control tests for individual SNPs using a standard allelic test with 1 df. In the next step, after having ascertained which alleles from the SNPs identified in GWA studies 1-4 were carrying the risk for higher BMI, we combined the information from 12 SNPs by counting the number of BMI-increasing alleles present in each subject. Only subjects with 12 complete genotypes were included, i.e. 225 cases and 1351 controls. This score was entered into a logistic regression model with case-control status as an outcome.
20 18 16 14 12 cases 10 controls 8 6 4
Percentage of individuals _ individuals of Percentage 2 0 <=7 8 9 10 11 12 13 14 15 16 17 >=18 Number of BMI increasing alleles
Figure 1. Distribution of BMI-increasing alleles in cases and controls.
Results
The analysis of individual SNPs showed an absence of association between any of the studied markers and AN (allelic test, 1 df). Assuming a power of 80%, an α-level of .05 and a disease prevalence of 2.2 % 13, we would be able to detect an association with odds ratio of at least 1.3 or 0.77 for a heterozygote and 1.7 or 0.59 for a risk homozygote for most of the SNPs
40 Chapter 2 individually, except for rs6548238, rs1488830, rs7647305 for which the effect sizes would have to be larger (Table I). Performing the same analysis solely on the ANR subset yielded nearly identical results, but with diminished power. For this reason both subsets are taken together in the study. To make sure that obese individuals in the control group were not relevantly influencing results we conducted a separate single SNP analysis with obese controls (BMI>30) excluded. Results were not materially different (data not shown). To test whether variants increasing BMI in the general population play a role in AN, we entered the combined number of effect alleles (i.e. BMI-increasing alleles from GWAS 1-4) into a logistic regression model. Frequencies of cases and controls per number of effect alleles are shown in Fig. 1 and the results of the logistic regression analysis are presented in Table II. The logistic regression analysis, assuming an α-level of .05, 80% power and a two-sided hypothesis, would be able to detect a change from the baseline probability (prevalence of the disease) of .02 to .05 with an increase of one SD (SD=2.47) in a number of effect alleles.
TABLE II. Logistic regression: the number of effect alleles is not associated with probability of being a case
95% CI p- Independent variable Β-coefficient df OR value Lower Upper
Number of effect alleles .00 1 .84 1.00 .95 1.06 Β-coefficient represents a change in probability of being a case with each additional risk- allele; OR, represents an increase in the odds of being a case with each additional risk-allele. The analysis included only individuals with 12 complete genotypes; n cases = 225; n controls = 1351. Mean (SD) number of effect alleles in cases = 12.20(2.32) and controls = 12.17(2.50). CI, confidence intervals.
The number of BMI-increasing alleles was not associated with the risk of AN (p = 0.84; Table II).
41 Chapter 2
Accordingly, mean numbers of effect alleles were similar between groups (p=.48 in a t-test) with a mean (SD) of 12.3 (2.4) alleles in cases and 12.2 (2.5) in controls (with α=.05 and at 80% power the test would detect difference in means of at least .18).
2.25
2
1.75
1.5
1.25
1
Odds ratio _ ratio Odds 0.75
0.5
0.25
0 <=8 9 10 11 12 13 14 15 >=16 Number of effect alleles
FIG 2. A plot showing OR for being a case along increasing number of effect alleles. The vertical bars represent 95% CIs.
Discussion
In the present study 12 SNPs found to be associated with BMI in the general population 1-4 were successfully genotyped in AN cases (n=267) and population controls (n=1636). We found no evidence for association with the risk of AN. Furthermore, by calculating the combined score of effect alleles (i.e. BMI-increasing alleles from GWA studies of BMI 1-4) we tested whether genetic variants increasing BMI in the general population play a role in AN. Logistic regression model revealed absence of association between the number of effect alleles and the risk of AN. These results contribute to the discussion about a supposed continuum of eating disorders, normal weight range and obesity 14,15. In our study, on the level of genetic aetiology, AN appeared to be a discrete entity rather than a part of this continuum.
42 Chapter 2
A limitation of the current study is its relatively small sample size and thus limited power to identify small effect sizes. Our sample has sufficient power (80%) to identify effect sizes of at least 1.3 OR at a 5% α-level, which is substantially larger than the effect sizes (OR 1.07 – 1.67) reported for obesity or extreme childhood obesity in the original GWA studies 1-4. With the same assumptions, a combined analysis of BMI-increasing alleles could detect a change in risk of the disease from baseline of .02 to .05, with an increase of one SD (2.47) in a number of effect alleles. To reduce phenotypic heterogeneity, we focused solely on AN because this subtype is distinct from the other types of eating disorders 15,16. Our main conclusions are based on the analysis of the combined score of effect alleles which, along with the fact that mean numbers of effect alleles between cases and controls were remarkably similar (power in the t-test sufficient to detect a difference of at least .18), shows that the investigated SNPs had no significant impact on susceptibility to AN. In this study no support was found for the hypothesis that the common genetic variants influencing BMI in the general population are substantial risk factors of AN, suggesting that effects of those variants may be overridden by other genetic factors of susceptibility to the disease. However, we cannot exclude that some association might be found with a considerable increase in sample size and refinement of phenotypes. In conclusion, this study found no evidence that SNPs which were previously proven to be robustly associated with BMI in the general population protect against or contribute to the risk of AN.
Acknowledgements
We are thankful to The GIANT (Genetic Investigation of ANtropometric Traits) Consortium for sharing the data on SNPs associated with BMI. This work was supported by funding from the Marie Curie Research Training Network INTACT (Individually tailored stepped care for women with eating disorders; reference number: MRTN-CT-2006-035988).
43 Chapter 2
We thank Bobby Koeleman, Behrooz Alizadeh and Caroline de Kovel for helpful comments.
Financial disclosures
The authors reported no potential conflicts of interests.
References
1. Willer CJ, Speliotes EK, Loos RJ, Li S, Lindgren CM, Heid IM, et al. Six new loci associated with body mass index highlight a neuronal influence on body weight regulation. Nat Genet 2009;41:25-34. 2. Thorleifsson G, Walters GB, Gudbjartsson DF, Steinthorsdottir V, Sulem P, Helgadottir A, et al. Genome-wide association yields new sequence variants at seven loci that associate with measures of obesity. Nat Genet 2009;41:18- 24. 3. Frayling TM, Timpson NJ, Weedon MN, Zeggini E, Freathy RM, Lindgren CM, et al. A common variant in the FTO gene is associated with body mass index and predisposes to childhood and adult obesity. Science 2007;316:889- 894. 4. Loos RJF, Lindgren CM, Li S, Wheeler E, Zhao JH, Prokopenko I, et al. Common variants near MC4R are associated with fat mass, weight and risk of obesity. Nat Genet 2008;40:768-775. 5. Stice E, Killen JD, Hayward C, Taylor CB. Support for the continuity hypothesis of bulimic pathology, J Consult Clin Psychol 1998;66:784-790. 6. Steinhausen H, Grigoroiu-Serbanescu M, Boyadjieva S, Neumärker K, Metzke CW. The relevance of body weight in the medium-term to long-term course of adolescent anorexia nervosa. findings from a multisite study. Int J Eat Disord 2008;9999:NA.
44 Chapter 2
7. Hebebrand J, Remschmidt H. Anorexia nervosa viewed as an extreme weight condition: Genetic implications. Hum Genet 1995;95:1-11. 8. Coners H, Remschmidt H, Hebebrand J. The relationship between premorbid body weight, weight loss, and weight at referral in adolescent patients with anorexia nervosa. Int J Eat Disord 1999;26:171-178. 9. Cooper Z, Fairburn C. The eating disorder examination: A semi-structured interview for the assessment of the specific psychopathology of eating disorders. Int J Eat Disord 1987;6:1-8. 10. Boker LK, van Noord PA, van der Schouw YT, Koot NV, Bueno de Mesquita HB, Riboli E, et al. Prospect-EPIC utrecht: Study design and characteristics of the cohort population. european prospective investigation into cancer and nutrition. Eur J Epidemiol 2001;17:1047-1053. 11. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MAR, Bender D, et al. PLINK: A tool set for whole-genome association and population-based linkage analyses. The American Journal of Human Genetics 2007;81:559-575. 12. Dudbridge F. Likelihood-based association analysis for nuclear families and unrelated subjects with missing genotype data. Hum Hered 2008;66:87- 98. 13. Keski-Rahkonen A, Hoek HW, Susser ES, Linna MS, Sihvola E, Raevuori A, et al. Epidemiology and course of anorexia nervosa in the community. Am J Psychiatry 2007;164:1259-1265. 14. Collier DA, Treasure JL. The aetiology of eating disorders. The British Journal of Psychiatry 2004;185:363-365. 15. Gleaves DH, Brown JD, Warren CS. The continuity/discontinuity models of eating disorders: A review of the literature and implications for assessment, treatment, and prevention. Behav Modif 2004;28:739-762. 16. Keel PK, Fichter M, Quadflieg N, Bulik CM, Baxter MG, Thornton L, et al. Application of a latent class analysis to empirically define eating disorder phenotypes. Arch Gen Psychiatry 2004;61:192-200.
45 Chapter 3
Chapter 3
Association study of POMC variants with body composition measures and nutrient choice
Andrew Ternouth* Marek K. Brandys* Yvonne T. van der Schouw Judith Hendriks John-Olov Jansson David Collier Roger A. Adan
*These authors contributed equally
European Journal of Pharmacology 2011; 660(1): 220-5.
46 Chapter 3
Abstract
Genome linkage scans and candidate gene studies have implicated the pro- opiomelanocortin (POMC) locus in traits related to food intake, metabolic function, and body mass index. Here we investigate single nucleotide polymorphisms at the POMC locus in order to evaluate the influence of its genetic variance on body fat distribution and diet in a sample of middle-aged men from the Netherlands. 366 Dutch males from the Hamlet cohort were asked detailed questions about food choice, nutrient intake and exercise. Furthermore, their weight and body fat composition were measured. Each cohort member was genotyped for a set of single nucleotide polymorphisms (SNPs) at the POMC locus. Regression analysis, adjusted for several covariates, was used to test for association between genetic variants and the phenotypes measured. POMC variation was associated with waist:hip ratio, visceral fat and abdominal fat (rs6713532, P=0.020, 0.019, 0.021, respectively), and nutrient choice (rs1042571, P=0.034), but in light of limited power and multiple testing these results should be taken with caution. POMC is a strong candidate for involvement in appetite regulation as supported by animal, physiological, and genetic studies and variation at the POMC locus may affect an individual’s energy intake which in turn leads to variation in body composition and body fat.
47 Chapter 3
1. Introduction
The brain melanocortin system has an established role in the physiology of weight regulation and consists of the pro-opiomelanocortin (POMC) gene, which encodes the melanocortins (MCs) and β-endorphin, the melanocortin receptor genes and the Agouti-related protein (AgRP) gene. POMC, MC3 and
MC4 receptor knockout mice are all obese. In addition to obesity, POMC deficient mice display strongly increased weight gain on a high fat diet, while eating only a modest amount more than wild-type mice (Yaswen et al., 1999). Genetic as well as pharmacological studies in rodents indicate that reduction of MC receptor activity is associated with increased fat intake on choice diets (Hagan et al., 2001; Koegler et al., 1999). Activation of the MC system results in reduction of fat intake, an effect which is dependent on the
MC4 receptor (Samama et al., 2003). This implication in food intake and fat deposition, coming from rodent studies, makes genes belonging to the melanocortin system good candidates for human association studies with nutrient choice and anthropometric measures. Indeed, the melanocortin system is implicated in the development of obesity in humans, as mutations in several of its component genes are strongly associated with rare, penetrant, monogenic forms of obesity (Farooqi, 2006). For example heterozygous mutations in the melanocortin-4 receptor gene are the most common monogenic form of severe obesity in children, affecting about 2.6% of the population, and mis-sense mutations in POMC can result in monogenic sever early-onset obesity (Farooqi et al., 2006; Challis et al., 2002; Krude et al., 1998).
A common MC4 receptor polymorphism has been associated with body mass index (BMI) (Heid et al., 2005) and a common variation in the AgRP gene has been associated with leanness and fat intake (Loos et al., 2005; Bonilla et al., 2006; Marks et al., 2004). For POMC itself, linkage studies have implicated its chromosomal locus at 2p23 in nutrient intake (Cai et al., 2004), fat mass (Comuzzie et al., 1997), and obesity (Hager et al., 1998). Other related phenotypes such as blood pressure (Rice et al., 2002), leptin
48 Chapter 3 levels (Suviolahti et al., 2003), physical activity (Simonen et al., 2003), and metabolic syndrome (Loos et al., 2003) have also been linked to the POMC region. Other studies found association of the POMC gene with BMI, weight, and total fat (Chen et al., 2005) (in a female population), BMI, waist:hip ratio, subcutaneous fat, and visceral fat (Sutton et al., 2005) (in a Hispanic population), and waist:hip ratio (in a general UK population) (Baker et al., 2005). There have also been negative reports which do not show evidence for linkage or association with BMI or related phenotypes at the POMC locus (Suviolahti et al., 2003; Delplanque et al., 2000). There has been no attempt to associate macronutrient intake with genetic variation in the POMC gene. The strength of the current study is the detailed measurement of body composition and macronutrient intake phenotypes available in the Hamlet cohort, which may help to overcome some of the difficulties potentially arising from the use of indirect phenotypes such as BMI. The present study was designed to replicate associations between single nucleotide polymorphisms at the POMC locus, obesity phenotypes (BMI, total fat mass and waist:hip ratio) and measures of macronutrient intake.
2. Materials and Methods
2.1 Subjects We conducted a population-based cross-sectional, single-centre study among 400 men aged between 40 and 80 years and living independently. The subjects and methods of recruitment have been described elsewhere (Muller et al., 2005). Data collection took place during two interviews in which data was self reported or measured by a trained clinician. All participants gave written informed consent before enrolment in the study, and the study was approved by the institutional review board of the University Medical Center Utrecht. Data were collected between March 2001 and April 2002.
49 Chapter 3
2.2 Phenotypes 2.2.1 BMI Height and weight were measured in standing position without shoes. BMI was calculated as the weight in kilograms divided by the square of the height in metres. 2.2.2 Waist:Hip Ratio Waist circumference was measured midway between the lower rib margin and iliac crest with subject in a standing position. Hip circumference was measured in the same standing position at the level of the greater trochanter. Each reading was taken twice and the mean used for calculation of the ratio. 2.2.3 Visceral and Subcutaneous Fat Visceral fat were measured using ultrasonography (Stolk et al., 2001) with an HDI 3000 (Philips Medical Systems, Eindhoven, the Netherlands) using a C 4- 2 transducer. The distances between the posterior edge of the abdominal muscles and the lumbar spine or psoas muscles are measured using electronical callipers. For all images the transducer is placed on a straight line drawn between the left and right midpoint of lower rib and iliac crest. Distances are measured from three different angles: medial, left and right for intraabdominal fat mass and medial for subcutaneous fat mass in threefold. Measurements are made at the end of quiet expiration, applying minimal pressure without displacement of intraabdominal contents as observed by ultrasound image. Visceral fat was measured as the distance between the skin and the linea alba and intraabdominal fat as the distance between the peritoneum and lumbar spine. Abdominal fat was calculated as the sum of visceral and subcutaneous fat. 2.2.4 Total Fat Mass, Lean Mass Total and trunk lean body mass, fat mass were measured using dual-energy x-ray absorptiometry (Hologic QDR 1000 densitometer, Hologic Inc., Waltham, MA, USA). Quality assurance for dual-energy x-ray absorptiometry,
50 Chapter 3 including calibration, was performed every morning, using the standard provided by the manufacturer. 2.2.5 Physical Exercise During interview, subjects completed Voorrips questionnaire (Voorrips et al., 1991) which provides a validated reliable measure of physical activity. 2.2.6 Nutrient Intake A validated food frequency questionnaire (FFQ) was administered, designed to estimate regular intake of 178 food items in the year before enrolment (Ocke et al., 1997a; Ocke et al., 1997b). From this questionnaire we calculated daily total energy, protein, fat, carbohydrate, and alcohol intake. From these data we additionally calculated fat:protein, fat:carbohydrate, and carbohydrate:protein ratios.
2.3 Choice of genetic markers Haplotype-tagging SNPs were selected in the coding sequence and up- and downstream of the POMC gene (UCSC browser coordinates chr2:25,292,212- 25,303,950, May 2004 freeze), using HapMap (Haploview 3.32) with r2 threshold set at > 0.80, using data available at time of study design (December 2006). Four SNPs were used to tag the POMC locus, although only three (rs6713532, rs6545975, and rs934778) were genotyped because a Taqman assay could not be designed for rs7565877. Secondly, a literature search was conducted to investigate POMC SNPs which had already been associated with feeding or fat related phenotypes with a minor allele frequency>0.05 (Cai et al., 2004; Comuzzie et al., 1997; Hager et al., 1998; Baker et al., 2005). Three SNPs were chosen: rs1009388, rs1042571, and rs1866146 (Supplementary Fig. 1).
2.4 Genotyping Genomic DNA had previously been isolated from peripheral lymphocytes using a high salt extraction procedure. DNA was not available for two men, they were therefore excluded.
51 Chapter 3
All genotyping was performed using Taqman assay by design or assay on demand genotyping kits (http://www.appliedbiosystems.com) using standard protocol. Fluorescence intensities were quantified using an Applied Biosystems HT7900.
2.5 Statistical Analysis Phenotypes were tested for normality of distribution and, where necessary, logarithmic (natural) or square-root transformations were applied. SNPs were tested for violation of the Hardy-Weinberg equilibrium with the exact test (Wigginton et al., 2005). Association of SNPs with quantitative traits was tested via linear regression with adjustment for several covariates, assuming an additive model of genetic effect (allele dosage). Analyses were conducted with Plink (Purcell et al., 2007) and SPSS 15.0 (SPSS, Chicago, Illinois).
3. Results
General characteristics of the subjects are presented in Table 1.
Table 1 Descriptive statistics detailing demographic, body composition and feeding phenotypes. Standard Phenotype Na Mean Median Minimum Maximum Deviation Demographic Age (years) 366 60.53 61.00 11 40 80 Body Fat and Composition bmi (kg/m2) 366 26.26 26.14 3.38 17.29 43.31 waist:hip ratio 366 0.98 0.98 0.06 0.77 1.17 visceral fat (cm) 366 7.53 7.10 2.18 3.3 14.75 subcutaneous fat (cm) 366 2.65 2.60 0.86 0.3 6.5 abdominal fat (cm) 366 10.18 9.78 2.43 4.95 20.45 total fat mass (kg) 346 17.1 16.23 5.48 5.91 49.33 total lean mass (kg) 346 61.52 61.34 7.17 44.81 88.77 Energy Balance
52 Chapter 3
physical activity 363 18.08 17.67 7.47 1.1 43.95 total energy intake (kcal) 364 2239.81 2216.28 533.36 735.79 4276.01 protein intake (g/day) 364 83.63 81.98 20.73 38.06 162.46 fat intake (g/day) 364 89.51 87.29 27.37 29.71 193.53 carbohydrate intake (g/day) 364 241.39 232.38 66.91 69.75 605.24 alcohol intake (g/day) 364 19.15 13.56 19.48 0 153.37 Nutrient Choice fat:protein ratio 364 1.07 1.07 0.19 0.61 1.75 fat:carbohdrate ratio 364 0.38 0.37 0.1 0.13 0.79 carbohydrate:protein ratio 364 2.94 2.88 0.66 1.38 5.94 a Sample consisted of males
All 6 SNPs were in Hardy-Weinberg equilibrium (exact test, (Wigginton et al., 2005)) (Table 2). 32 out of 398 subjects were excluded due to low genotyping rate (more than 2 missing genotypes). Total genotyping rate in the remaining individuals was 0.89.
Table 2 Genotype distributions and Hardy-Weinberg equilibrium.
Polymorphism Genotype AA Genotype AB Genotype BB Hardy-Weinberg SNP (minor allele first) count (%) count (%) count (%) P (exact test) rs1866146 G/A 43 (12.1%) 167 (47.0%) 145 (40.8%) 0.73 rs1042571 T/C 9 (2.8%) 80 (24.7%) 234 (72.4%) 0.51 rs6713532 C/T 18 (5.1%) 129 (36.5%) 206 (58.3%) 0.77 rs6545975 C/T 58 (18.4%) 143 (45.5%) 113 (35.9%) 0.30 rs934778 G/A 4 (1.5%) 37 (13.7%) 229 (84.8%) 0.10 rs1009388 G/C 26 (7.7%) 122 (36.3%) 188 (55.9%) 0.32
The successfully genotyped SNPs were re-evaluated using the Tagger program (http://www.broad.mit.edu/mpg/tagger/) (de Bakker et al., 2005) to estimate the amount of genetic variation captured. Applying a r2 threshold of 0.8 and using the pairwise method, 4 HapMap SNPs genotyped in this study cover about 57% of the genetic variation on the POMC gene locus (43% when the region is extended by 5kb at 5’ and 3’ ends). One Hapmap SNP had been excluded from analysis as no Taqman assay was available. Two genotyped SNPs (rs1042571 and rs1009388) were not in Hapmap. Their
53 Chapter 3 presence increases the extent of captured variance but it is difficult to quantify it. Table 3 displays r2 values between the SNPs. The highest r2 value was 0.48 for rs1042571 and rs1009388. The values in this dataset are generally low or very low, meaning that each SNP contributes independent genetic information.
Table 3 r2 as a measure of linkage disequilibrium. SNP rs1866146 rs1042571 rs6713532 rs6545975 rs934778 rs1009388 rs1866146 - 0.10 0.38 0.00 0.04 0.17 rs1042571 - 0.03 0.06 0.10 0.48 rs6713532 - 0.11 0.00 0.09 rs6545975 - 0.00 0.15 rs934778 - 0.04 rs1009388 -
Results of the genetic association analysis are shown in Table 4, which displays P values for association, as determined by linear regression modelling with adjustment for age, education, physical activity, alcohol intake and BMI. The strongest associations were found for rs6713532 and waist:hip ratio, visceral fat and abdominal fat (i.e. sum of visceral and subcutaneous fat). These associations became weaker when not controlled for BMI. SNP rs1042571 was associated with a ratio between fat and protein content in the diet. Furthermore, 2 other SNPs (rs6545975, rs1009388) displayed suggestive association with investigated phenotypes (Table 4).
Table 4 P values < 0.2 from the linear regression modelling of the association between POMC variants and quantitative phenotypes (additive model). Adjusted for age, education, physical activity, alcohol intake and BMI (BMI as a covariate excluded when testing for association with BMI). Phenotype / SNP rs1866146 rs1042571 rs6713532 rs6545975 rs934778 rs1009388 Body Fat and
Composition
54 Chapter 3
BMI (kg/m2)a 0.104 waist:hip ratio 0.020 0.061 0.141 visceral fat (cm)a 0.019 0.090 0.101 subcutaneous fat (cm) 0.078 abdominal fat (cm)a 0.021 0.133 total fat mass (kg) total lean mass (kg)a 0.115 0.167 Energy Balance total energy intake 0.196 0.177 (kcal)a protein intake (g/day)b 0.183 0.118 fat intake (g/day)a 0.080 carbohydrate intake 0.087 (g/day)b Nutrient Choice fat:protein ratioa 0.034 0.147 0.063 fat:carbohdrate ratioa 0.136 0.125 carbohydrate:protein 0.166 ratioa a natural log-transformed; b square-root transformed; P<0.1 in italics; P<0.05 in bold
Interestingly, when for exploratory purposes an additional covariate – total lean mass – was controlled for, the three significant association signals of rs6713532 became slightly more pronounced and two association signals for rs6545975 became significant (with waist:hip ratio, P=0.043, and visceral fat, P=0.049; Supplementary Table 2). Table 5 presents means and SDs of phenotypes per each genotype for SNPs significantly associated in a linear regression model, together with beta-coefficients.
Table 5 Means and SDs of the phenotype per each genotype, only for significant associations. mean (SD) mean (SD) mean (SD) for the for the beta- variance for the P major minor coefficient explained heterozygote homozygote homozygote
unstan- stan- Phenotype rs6713532 dardized dardized
55 Chapter 3
waist:hip 0.98 (0.06) 0.97 (0.06) -0.014 -0.09 0.82% 0.020 ratio 0.94 (0.05) visceral fat 7.76 (2.29) 7.41 (2.06) -0.065 -0.089 0.79% 0.019 (cm)a 6.55 (1.63) abdominal 10.4 (2.59) 10.08 (2.24) -0.060 -0.078 0.61% 0.021 fat (cm)a 8.96 (1.82) Phenotype rs1042571 fat:protein 1.06 (0.19) 1.08 (0.18) 1.23 (0.2) 0.071 0.119 1.41% 0.034 ratioa a log-transformed variables were used. For log-transformed variables, one minor allele increase is associated with an average of (100 x unstandardized beta) percent increase in the phenotype. For the non-transformed variable (waist:hip ratio), unstandardized beta-coefficient represents an averaged increase in the phenotype with each additional minor allele (mean for the major homozygote represents the intercept).
Furthermore, SNPs rs6713532 and rs1866146 were tested in a sample of 938 young men (mean age=18.9) from the GOOD study (genotyped on Human610K Illumina array) (Andersson et al., 2009). Available phenotypes allowed testing for association with BMI and total fat mass (both log-transformed), adjusted for age. There was no association with either of the phenotypes (Supplementary Table 1). P values in this study are uncorrected for multiple testing. For a given sample size (n=366), in order to have 80% statistical power to detect a true association, a SNP associated with a quantitative trait would have to explain about 2% of the total variance of the trait (assuming alfa at 0.05 and a minor allele frequency of 25%), testing for additive effects of the quantitative trait locus only.
4. Discussion
This study should be viewed in light of several limitations. P values are uncorrected for multiple testing, and it is clear that none would ‘survive’ such a correction (even though some of the tested phenotypes are highly correlated; Supplementary Table 1). Furthermore, our statistical power was not sufficient to detect minor genetic effects, if such exist. Also a genotyping rate of 90% and an incomplete coverage of genetic variation on the POMC
56 Chapter 3 locus are limitations. For the reasons above, and since case-only studies are particularly prone to spurious genetic associations (Sullivan, 2007), the current results should be taken with caution. We have controlled for similar covariates as in Baker et al. (2005) (Baker et al., 2005), except for smoking, for which no data were available. The association of rs1042571 and rs1009388 with waist:hip ratio reported by Baker et al. (2005)(Baker et al., 2005) was not confirmed in our study (SNP rs1009388 has also been tested in (Bienertova-Vasku et al., 2010) where no association with BMI or waist:hip ratio has been found). SNP rs1009388 had suggestive P values in tests with fat intake and fat:protein ratio and rs1042571 was associated with fat:protein ratio, indicating that effects of those variants may be stronger on food choice than on body composition. It should be noted that the sample in the current study was different from Baker et al. with respect to sex (males only), age and body composition measures. We were able to test two SNPs from this study (rs6713532, rs1866146) in an independent sample of 938 young men. There was no association with either BMI or total fat mass. It was not possible to test for association with abdominal or visceral fat or waist:hip ratio – phenotypes that displayed strongest signals in the current study. This shows that the association of rs6713532 with measures of fat and waist:hip ratio is either a false-positive or it is highly specific to those phenotypes or studied population. The strength of the Hamlet cohort lies in the availability of a number of potentially relevant phenotypes. Studies have suggested that measurement of waist circumference or waist:hip ratio, as indicators of abdominal obesity, may be better disease risk predictors than the BMI (Yusuf et al., 2005; Wang et al., 2005; Noble, 2001). Moreover, visceral fat levels are more strongly related to poor outcome than subcutaneous fat levels (Albu et al., 2000). The current study shows that the waist:hip ratio, abdominal fat and visceral fat may be better measures for genetic studies of obesity and body composition than the commonly used BMI, which is more likely
57 Chapter 3 influenced by factors other than the fat tissue, such as muscularity or bone structure. This is supported by our exploratory analysis with additional adjustment for total lean mass, which resulted in augmentation of the association signals.
Association of SNPs near to the MC4 receptor gene are one of the most robust findings in genetic studies of body composition (Loos et al., 2008), (Chambers et al., 2008; Willer et al., 2009). POMC is located upstream of MC4 in a genetic pathway and our results further support the involvement of the common variation in the melanocortin system in regulation of fat mass in humans, additionally suggesting that the effects on body composition may be mediated by the association with food choice and nutritional content of the diet. This study also shows that research could benefit from investigating phenotypes alternative to BMI (such as waist:hip ratio) and adjusting for confounders which are likely to obfuscate the relation between genetic variants and levels of fatness.
Acknowledgements
Andrew Ternouth was supported by a Medical Research Council studentship. Marek K. Brandys is funded by the Marie Curie Research Training Network INTACT (Individually tailored stepped care for women with eating disorders; reference number: MRTN-CT-2006-035988). Roger A.H. Adan is funded by VIDI grant 016.036.322 from the Netherlands Organisation for Scientific Research (NWO).
References
Albu, J.B., Kovera, A.J., Johnson, J.A., 2000. Fat distribution and health in obesity. Ann. N. Y. Acad. Sci. 904, 491-501.
58 Chapter 3
Andersson, N., Strandberg, L., Nilsson, S., Ljungren, O., Karlsson, M.K., Mellstrom, D., Lorentzon, M., Ohlsson, C., Jansson, J., 2009. Variants of the interleukin-1 receptor antagonist gene are associated with fat mass in men. Int J Obes (Lond) 33, 525-533. Baker, M., Gaukrodger, N., Mayosi, B.M., Imrie, H., Farrall, M., Watkins, H., Connell, J.M.C., Avery, P.J., Keavney, B., 2005. Association between common polymorphisms of the proopiomelanocortin gene and body fat distribution: a family study. Diabetes 54, 2492-2496. Bienertova-Vasku, J., Bienert, P., Forejt, M., Tomandl, J., Brazdova, Z., Vasku, A., 2010. Genotype x nutrient association of common polymorphisms in obesity-related genes with food preferences and time structure of energy intake. Br. J. Nutr. 103, 352-359. Bonilla, C., Panguluri, R.K., Taliaferro-Smith, L., Argyropoulos, G., Chen, G., Adeyemo, A.A., Amoah, A., Owusu, S., Acheampong, J., Agyenim-Boateng, K., Eghan, B.A., Oli, J., Okafor, G., Abbiyesuku, F., Johnson, T., Rufus, T., Fasanmade, O., Chen, Y., Collins, F.S., Dunston, G.M., Rotimi, C., Kittles, R.A., 2006. Agouti-related protein promoter variant associated with leanness and decreased risk for diabetes in West Africans. Int J Obes (Lond) 30, 715-721. Cai, G., Cole, S.A., Bastarrachea, R.A., Bastarrachea-Sosa, R.A., Maccluer, J.W., Blangero, J., Comuzzie, A.G., 2004. Quantitative trait locus determining dietary macronutrient intakes is located on human chromosome 2p22. Am. J. Clin. Nutr. 80, 1410-1414. Challis, B.G., Pritchard, L.E., Creemers, J.W.M., Delplanque, J., Keogh, J.M., Luan, J., Wareham, N.J., Yeo, G.S.H., Bhattacharyya, S., Froguel, P., White, A., Farooqi, I.S., O'Rahilly, S., 2002. A missense mutation disrupting a dibasic prohormone processing site in pro-opiomelanocortin (POMC) increases susceptibility to early-onset obesity through a novel molecular mechanism. Hum. Mol. Genet. 11, 1997-2004. Chambers, J.C., Elliott, P., Zabaneh, D., Zhang, W., Li, Y., Froguel, P., Balding, D., Scott, J., Kooner, J.S., 2008. Common genetic variation near MC4R is associated with waist circumference and insulin resistance. Nat. Genet. 40, 716-718.
59 Chapter 3
Chen, Y., Snieder, H., Wang, X., Kaviya, B., McCaffrey, C., Spector, T.D., Carter, N.D., O'Dell, S.D., 2005. Proopiomelanocortin gene variants are associated with serum leptin and body fat in a normal female population. Eur. J. Hum. Genet. 13, 772-780. Comuzzie, A.G., Hixson, J.E., Almasy, L., Mitchell, B.D., Mahaney, M.C., Dyer, T.D., Stern, M.P., MacCluer, J.W., Blangero, J., 1997. A major quantitative trait locus determining serum leptin levels and fat mass is located on human chromosome 2. Nat. Genet. 15, 273-276. de Bakker, P.I.W., Yelensky, R., Pe'er, I., Gabriel, S.B., Daly, M.J., Altshuler, D., 2005. Efficiency and power in genetic association studies. Nat. Genet. 37, 1217-1223. Delplanque, J., Barat-Houari, M., Dina, C., Gallina, P., Clement, K., Guy-Grand, B., Vasseur, F., Boutin, P., Froguel, P., 2000. Linkage and association studies between the proopiomelanocortin (POMC) gene and obesity in caucasian families. Diabetologia 43, 1554-1557. Farooqi, I.S., 2006. Monogenic human obesity syndromes. Prog. Brain Res. 153, 119-125. Farooqi, I.S., Drop, S., Clements, A., Keogh, J.M., Biernacka, J., Lowenbein, S., Challis, B.G., O'Rahilly, S., 2006. Heterozygosity for a POMC-null mutation and increased obesity risk in humans. Diabetes 55, 2549-2553. Hagan, M.M., Rushing, P.A., Benoit, S.C., Woods, S.C., Seeley, R.J., 2001. Opioid receptor involvement in the effect of AgRP- (83-132) on food intake and food selection. Am. J. Physiol. Regul. Integr. Comp. Physiol. 280, 814- 821. Hager, J., Dina, C., Francke, S., Dubois, S., Houari, M., Vatin, V., Vaillant, E., Lorentz, N., Basdevant, A., Clement, K., Guy-Grand, B., Froguel, P., 1998. A genome-wide scan for human obesity genes reveals a major susceptibility locus on chromosome 10. Nat. Genet. 20, 304-308. Heid, I.M., Vollmert, C., Hinney, A., Doring, A., Geller, F., Lowel, H., Wichmann, H., Illig, T., Hebebrand, J., Kronenberg, F., 2005. Association of the 103I MC4R allele with decreased body mass in 7937 participants of two population based surveys. J. Med. Genet. 42.
60 Chapter 3
Koegler, F.H., Schaffhauser, R.O., Mynatt, R.L., York, D.A., Bray, G.A., 1999. Macronutrient diet intake of the lethal yellow agouti (Ay/a) mouse. Physiol. Behav. 67, 809-812. Krude, H., Biebermann, H., Luck, W., Horn, R., Brabant, G., Gruters, A., 1998. Severe early-onset obesity, adrenal insufficiency and red hair pigmentation caused by POMC mutations in humans. Nat. Genet. 19, 155-157. Loos, R.J.F., Lindgren, C.M., Li, S., Wheeler, E., Zhao, J.H., Prokopenko, I., Inouye, M., Freathy, R.M., Attwood, A.P., Beckmann, J.S., Berndt, S.I., Bergmann, S., Bennett, A.J., Bingham, S.A., Bochud, M., Brown, M., Cauchi, S., Connell, J.M., Cooper, C., Smith, G.D., Day, I., Dina, C., De, S., Dermitzakis, E.T., Doney, A.S.F., Elliott, K.S., Elliott, P., Evans, D.M., Sadaf Farooqi, I., Froguel, P., Ghori, J., Groves, C.J., Gwilliam, R., Hadley, D., Hall, A.S., Hattersley, A.T., Hebebrand, J., Heid, I.M., Herrera, B., Hinney, A., Hunt, S.E., Jarvelin, M., Johnson, T., Jolley, J.D.M., Karpe, F., Keniry, A., Khaw, K., Luben, R.N., Mangino, M., Marchini, J., McArdle, W.L., McGinnis, R., Meyre, D., Munroe, P.B., Morris, A.D., Ness, A.R., Neville, M.J., Nica, A.C., Ong, K.K., O'Rahilly, S., Owen, K.R., Palmer, C.N.A., Papadakis, K., Potter, S., Pouta, A., Qi, L., Randall, J.C., Rayner, N.W., Ring, S.M., Sandhu, M.S., Scherag, A., Sims, M.A., Song, K., Soranzo, N., Speliotes, E.K., Syddall, H.E., Teichmann, S.A., Timpson, N.J., Tobias, J.H., Uda, M., Ganz Vogel, C.I., Wallace, C., Waterworth, D.M., Weedon, M.N., Willer, C.J., Wraight, V.L., Yuan, X., Zeggini, E., Hirschhorn, J.N., Strachan, D.P., Ouwehand, W.H., Caulfield, M.J., Samani, N.J., Frayling, T.M., Vollenweider, P., Waeber, G., Mooser, V., Deloukas, P., McCarthy, M.I., Wareham, N.J., Barroso, I., 2008. Common variants near MC4R are associated with fat mass, weight and risk of obesity. Nat. Genet. 40, 768-775. Loos, R.J.F., Rankinen, T., Rice, T., Rao, D.C., Leon, A.S., Skinner, J.S., Bouchard, C., Argyropoulos, G., 2005. Two ethnic-specific polymorphisms in the human Agouti-related protein gene are associated with macronutrient intake. Am. J. Clin. Nutr. 82, 1097-1101. Loos, R.J.F., Katzmarzyk, P.T., Rao, D.C., Rice, T., Leon, A.S., Skinner, J.S., Wilmore, J.H., Rankinen, T., Bouchard, C., 2003. Genome-wide linkage scan
61 Chapter 3 for the metabolic syndrome in the HERITAGE Family Study. J. Clin. Endocrinol. Metab. 88, 5935-5943. Marks, D.L., Boucher, N., Lanouette, C., Perusse, L., Brookhart, G., Comuzzie, A.G., Chagnon, Y.C., Cone, R.D., 2004. Ala67Thr polymorphism in the Agouti- related peptide gene is associated with inherited leanness in humans. Am J Med Genet A 126A, 267-271. Muller, M., Grobbee, D.E., den Tonkelaar, I., Lamberts, S.W.J., van, d.S., 2005. Endogenous sex hormones and metabolic syndrome in aging men. J. Clin. Endocrinol. Metab. 90, 2618-2623. Noble, R.E., 2001. Waist-to-hip ratio versus BMI as predictors of cardiac risk in obese adult women. West. J. Med. 174, 240-241. Ocke, M.C., Bueno-de-Mesquita, H.B., Goddijn, H.E., Jansen, A., Pols, M.A., van Staveren, W.A., Kromhout, D., 1997a. The Dutch EPIC food frequency questionnaire. I. Description of the questionnaire, and relative validity and reproducibility for food groups. Int. J. Epidemiol. 26 Suppl 1, 37-48. Ocke, M.C., Bueno-de-Mesquita, H.B., Pols, M.A., Smit, H.A., van Staveren, W.A., Kromhout, D., 1997b. The Dutch EPIC food frequency questionnaire. II. Relative validity and reproducibility for nutrients. Int. J. Epidemiol. 26 Suppl 1, 49-58. Purcell, S., Neale, B., Todd-Brown, K., Thomas, L., Ferreira, M.A.R., Bender, D., Maller, J., Sklar, P., de Bakker, P.I.W., Daly, M.J., Sham, P.C., 2007. PLINK: A Tool Set for Whole-Genome Association and Population-Based Linkage Analyses. The American Journal of Human Genetics 81, 559-575 doi: DOI: 10.1086/519795. Rice, T., Rankinen, T., Chagnon, Y.C., Province, M.A., Perusse, L., Leon, A.S., Skinner, J.S., Wilmore, J.H., Bouchard, C., Rao, D.C., 2002. Genomewide linkage scan of resting blood pressure: HERITAGE Family Study. Health, Risk Factors, Exercise Training, and Genetics. Hypertension 39, 1037-1043. Samama, P., Rumennik, L., Grippo, J.F., 2003. The melanocortin receptor MCR4 controls fat consumption. Regul. Pept. 113, 85-88.
62 Chapter 3
Simonen, R.L., Rankinen, T., Perusse, L., Rice, T., Rao, D.C., Chagnon, Y., Bouchard, C., 2003. Genome-wide linkage scan for physical activity levels in the Quebec Family study. Med. Sci. Sports Exerc. 35, 1355-1359. Stolk, R.P., Wink, O., Zelissen, P.M., Meijer, R., van Gils, A.P., Grobbee, D.E., 2001. Validity and reproducibility of ultrasonography for the measurement of intra-abdominal adipose tissue. Int. J. Obes. Relat. Metab. Disord. 25, 1346- 1351. Sullivan, P.F., 2007. Spurious Genetic Associations. Biol. Psychiatry 61, 1121- 1126 doi: DOI: 10.1016/j.biopsych.2006.11.010. Sutton, B.S., Langefeld, C.D., Williams, A.H., Norris, J.M., Saad, M.F., Haffner, S.M., Bowden, D.W., 2005. Association of proopiomelanocortin gene polymorphisms with obesity in the IRAS family study. Obes. Res. 13, 1491- 1498. Suviolahti, E., Ridderstrale, M., Almgren, P., Klannemark, M., Melander, O., Carlsson, E., Carlsson, M., Hedenbro, J., Orho-Melander, M., 2003. Pro- opiomelanocortin gene is associated with serum leptin levels in lean but not in obese individuals. Int. J. Obes. Relat. Metab. Disord. 27, 1204-1211. Voorrips, L.E., Ravelli, A.C., Dongelmans, P.C., Deurenberg, P., Van Staveren, W.A., 1991. A physical activity questionnaire for the elderly. Med. Sci. Sports Exerc. 23, 974-979. Wang, Y., Rimm, E.B., Stampfer, M.J., Willett, W.C., Hu, F.B., 2005. Comparison of abdominal adiposity and overall obesity in predicting risk of type 2 diabetes among men. Am. J. Clin. Nutr. 81, 555-563. Wigginton, J.E., Cutler, D.J., Abecasis, G.R., 2005. A note on exact tests of Hardy-Weinberg equilibrium. Am. J. Hum. Genet. 76, 887-893. Willer, C.J., Speliotes, E.K., Loos, R.J., Li, S., Lindgren, C.M., Heid, I.M., Berndt, S.I., Elliott, A.L., Jackson, A.U., Lamina, C., Lettre, G., Lim, N., Lyon, H.N., McCarroll, S.A., Papadakis, K., Qi, L., Randall, J.C., Roccasecca, R.M., Sanna, S., Scheet, P., Weedon, M.N., Wheeler, E., Zhao, J.H., Jacobs, L.C., Prokopenko, I., Soranzo, N., Tanaka, T., Timpson, N.J., Almgren, P., Bennett, A., Bergman, R.N., Bingham, S.A., Bonnycastle, L.L., Brown, M., Burtt, N.P., Chines, P., Coin, L., Collins, F.S., Connell, J.M., Cooper, C., Smith, G.D.,
63 Chapter 3
Dennison, E.M., Deodhar, P., Elliott, P., Erdos, M.R., Estrada, K., Evans, D.M., Gianniny, L., Gieger, C., Gillson, C.J., Guiducci, C., Hackett, R., Hadley, D., Hall, A.S., Havulinna, A.S., Hebebrand, J., Hofman, A., Isomaa, B., Jacobs, K.B., Johnson, T., Jousilahti, P., Jovanovic, Z., Khaw, K.T., Kraft, P., Kuokkanen, M., Kuusisto, J., Laitinen, J., Lakatta, E.G., Luan, J., Luben, R.N., Mangino, M., McArdle, W.L., Meitinger, T., Mulas, A., Munroe, P.B., Narisu, N., Ness, A.R., Northstone, K., O'Rahilly, S., Purmann, C., Rees, M.G., Ridderstrale, M., Ring, S.M., Rivadeneira, F., Ruokonen, A., Sandhu, M.S., Saramies, J., Scott, L.J., Scuteri, A., Silander, K., Sims, M.A., Song, K., Stephens, J., Stevens, S., Stringham, H.M., Tung, Y.C., Valle, T.T., Van Duijn, C.M., Vimaleswaran, K.S., Vollenweider, P., Waeber, G., Wallace, C., Watanabe, R.M., Waterworth, D.M., Watkins, N., Wellcome Trust Case Control Consortium, Witteman, J.C., Zeggini, E., Zhai, G., Zillikens, M.C., Altshuler, D., Caulfield, M.J., Chanock, S.J., Farooqi, I.S., Ferrucci, L., Guralnik, J.M., Hattersley, A.T., Hu, F.B., Jarvelin, M.R., Laakso, M., Mooser, V., Ong, K.K., Ouwehand, W.H., Salomaa, V., Samani, N.J., Spector, T.D., Tuomi, T., Tuomilehto, J., Uda, M., Uitterlinden, A.G., Wareham, N.J., Deloukas, P., Frayling, T.M., Groop, L.C., Hayes, R.B., Hunter, D.J., Mohlke, K.L., Peltonen, L., Schlessinger, D., Strachan, D.P., Wichmann, H.E., McCarthy, M.I., Boehnke, M., Barroso, I., Abecasis, G.R., Hirschhorn, J.N., Genetic Investigation of ANthropometric Traits Consortium, 2009. Six new loci associated with body mass index highlight a neuronal influence on body weight regulation. Nat. Genet. 41, 25-34 doi: 10.1038/ng.287. Yaswen, L., Diehl, N., Brennan, M.B., Hochgeschwender, U., 1999. Obesity in the mouse model of pro-opiomelanocortin deficiency responds to peripheral melanocortin. Nat. Med. 5, 1066-1070. Yusuf, S., Hawken, S., Ounpuu, S., Bautista, L., Franzosi, M.G., Commerford, P., Lang, C.C., Rumboldt, Z., Onen, C.L., Lisheng, L., Tanomsup, S., Wangai, P., Razak, F., Sharma, A.M., Anand, S.S., 2005. Obesity and the risk of myocardial infarction in 27,000 participants from 52 countries: a case-control study. Lancet 366, 1640-1649.
64 Chapter 3
Supplementary data
Supplementary Table 1 P values < 0.2 from the linear regression modelling of the association between POMC variants and quantitative phenotypes (additive model). Adjusted for age, education, physical activity, alcohol intake, total lean mass, and BMI (BMI as a covariate excluded when testing for association with BMI; total lean mass not tested).
Phenotype / SNP rs1866146 rs1042571 rs6713532 rs6545975 rs934778 rs1009388
Body Fat and
Composition BMI (kg/m2)a 0.098 waist:hip ratio 0.016 0.043 0.149 visceral fat (cm)a 0.013 0.049 0.107 subcutaneous fat 0.076 (cm) abdominal fat (cm)a 0.014 0.141 total fat mass (kg) total lean mass (kg)a Energy Balance total energy intake 0.188 0.168 (kcal)a protein intake 0.177 0.106 (g/day)b fat intake (g/day)a 0.070 carbohydrate intake 0.092 (g/day)b Nutrient Choice fat:protein ratioa 0.022 0.137 0.054 fat:carbohdrate 0.158 0.090 0.106 ratioa carbohydrate:protein ratioa a natural log-transformed; b square-root transformed; P<0.1 in italics; P<0.05 in bold
65 Chapter 3
Supplementary Table 2 SNPs tested for association with BMI and fat mass in a sample of young men from the GOOD study (additive model). Adjusted for age. Both variables are log-transformed.
SNP beta SE Chi-Sq P MAF call rate n BMI rs1866146 -0.002 0.003 0.707 0.400 0.35 100% 938 rs6713532 0 0.003 0.019 0.890 0.24 100% 938 Fat mass rs1866146 -0.008 0.011 0.507 0.476 0.35 100% 938 rs6713532 -0.008 0.012 0.419 0.517 0.24 100% 938
66 Chapter 3
Supplementary Table 3 Correlations between phenotypes.
Phenotype 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. - - - - 1. BMI (kg/m2)a - .546** .664** .443** .757** .842** .671** -.025 -.083 -.010 -.103* .133* .168** .137** .221** .240** waist:hip - - - - 2. - .711** .168** .712** .571** .165** -.127* -.090 -.004 .024 .142** ratio .204** .169** .241** .148** visceral fat - - - 3. - .098 .927** .683** .356** -.113* -.058 -.041 .049 .015 .198** (cm)a .226** .234** .221** subcutaneou - - 4. - .451** .507** .370** -.064 -.061 .061 -.047 -.083 -.033 .032 s fat (cm) .162** .170** abdominal fat - - - 5. - .798** .461** -.133* -.039 -.066 .028 -.052 .179** (cm)a .225** .244** .253** total fat mass - - - 6. - .572** -.104 -.002 -.029 .001 -.044 .195** (kg) .174** .224** .265** total lean - 7. - -.013 .143** .209** .116* .052 .051 -.099 .070 mass (kg)a .162** physical - 8. - .106* .064 .018 .166** .024 -.068 .130* activityb .156** total energy 9. - .849** .891** .816** .177** .320** .167** .065 intake (kcal)a protein intake - 10. - .805** .654** .018 -.027 .244** (g/day)b .309** fat intake - 11. - .598** .049 .566** .524** (g/day)a .155** carbs intake - 12. - -.202** .103* .510** (g/day)b .362** alcohol - 13. intake - .049 .269** .277** (g/day)b fat:protein 14. - .546** .172** ratioa fat:carbs - 15. - ratioa .731** carbs:protein 16. - ratioa
a natural log-transformed; b square root transformed; * correlation is significant at the 0.05 level (2-tailed); ** correlation is significant at the 0.01 level (2-tailed)
67 Chapter 3
Fig. 1. SNPs picked for association study at the POMC locus. Filled rectangles denote untranslated exonic sequence and hatched rectangles denote coding exonic sequence.
68 Chapter 4
Chapter 4
Anorexia nervosa and the Val158Met polymorphism of the COMT gene: meta-analysis and new data
Marek K. Brandys Margarita C.T. Slof-Op ’t Landt Annemarie A. van Elburg Roel Ophoff Willem Verduijn Ingrid Meulenbelt Christel M. Middeldorp Dorret I. Boomsma Eric F. van Furth P. Eline Slagboom Martien J. H. Kas Roger A. H. Adan
Psychiatric Genetics 2012; 22(3):130-6.
69 Chapter 4
Abstract
Objectives: This study aimed to test the association between the Val158Met polymorphism (rs4680) of the catechol-O-methyl transferase gene and anorexia nervosa. Methods: Firstly, an association study on two cohorts (306 cases and 1009 controls from Utrecht, and 174 cases and 466 controls from Leiden/NTR) was performed. Subsequently, the results were integrated into a meta-analysis, together with all the case-control and family-based studies, which were testing the same hypothesis and were available in the literature. Altogether, 8 studies (11 datasets) were included in this meta-analysis, with a total of 2021 cases, 2848 controls and 89 informative (heterozygous) trios. Results: The present association studies found no association between AN and rs4680 when testing the allelic contrast (Utrecht OR=1.14, P=0.14; Leiden OR=1.02, P=0.85). There was an indication of association under the dominant model of genetic effect in the Utrecht cohort (for the Met allele, OR=1.42, P=0.03). Nevertheless, the meta-analyses of both the allelic contrast and the dominant effect were non-significant (the allelic pooled OR=1.03, P= 0.42 and the dominant pooled OR = 1.1, P=0.18). The meta- analyses were performed under the fixed-effect model and there was no significant heterogeneity among the effect sizes. Conclusions: Meta-analytically combined evidence from the present genotypings and the literature search shows that the effect sizes are homogeneous across studies and that rs4680 is not associated with AN.
70 Chapter 4
Introduction
Anorexia nervosa (AN) is a debilitating disease, notorious for its highest standardized mortality ratio among all psychiatric illnesses (with 6 to 10 times higher mortality rate, compared to the general population (Birmingham, Su, Hlynsky, Goldner, & Gao, 2005; Papadopoulos, Ekbom, Brandt, & Ekselius, 2009)). Despite the seriousness of anorexia, its aetiology remains elusive. Several twin and adoption studies determined that genetic factors explain 46 to 78 percent of variance in AN (Bulik et al., 2010; Kortegaard, Hoerder, Joergensen, Gillberg, & Kyvik, 2001; Wade, Bulik, Neale, & Kendler, 2000). A family study showed a 10-fold increase in lifetime risk of AN for a first-degree female relative of a person affected by an eating disorder (ED) (comparing to relatives of unaffected individuals) (Strober, Freeman, Lampert, Diamond, & Kaye, 2000). So far studies on genetic risk factors of AN focused on candidate genes of neurotransmitter and neuropeptide pathways. The yield of those studies has been inconclusive and many promising findings have not been replicated (Pinheiro et al., 2010). One of the genes that were proposed to play a role in the development of AN susceptibility is catechol-O-methyl transferase (COMT). COMT is an enzyme responsible for degradation of catecholamines, such as dopamine and noradrenaline (Chen et al., 2004). COMT has been implicated in the pathogenesis of several mental disorders and in explaining variation in cognitive phenotypes (Mier, Kirsch, & Meyer-Lindenberg, 2010). In particular, rs4680 (Val/Met), a functional variant in the COMT locus, has been studied extensively. Rs4680 is located in exon 3 of the gene and results in a valine-to-methionine substitution. The Met allele of this gene has been associated with a less stable product and, therefore, lower enzymatic activity (Lachman et al., 1996), which in turn has been hypothesized to lead to higher dopamine availability (Shield, Thomae, Eckloff, Wieben, & Weinshilboum, 2004). Rs4680 was studied in mental disorders such as schizophrenia (Costas et al., 2011), autism (Stergiakouli& Thapar, 2010), depression (Åberg, Fandiño-Losada, Sjöholm, Forsell, & Lavebratt, 2011) and eating disorders
71 Chapter 4
(Frieling et al., 2006; Mikolajczyk, Smiarowska, Grzywacz, & Samochowiec, 2006). In the field of AN, initial encouraging findings (Frisch et al., 2001) were not replicated (Gabrovsek et al., 2004). However, the power to detect small effect sizes in those studies was limited. The aim of the present study was to test the hypothesis of association with more power and to provide a better estimation of the ES. To do so, we performed genotyping on 2 cohorts of patients with AN and controls from the Netherlands and combined the results in a meta-analysis of all the available case-control and family-based studies, which tested this same hypothesis. Not only is the statistical power of such an approach higher than in a single-cohort study, but it also provides insight into heterogeneity of effect sizes across the studies.
Methods
Association study Genotyping was performed on two sets of cases and controls from the Netherlands: the Utrecht cohort and the Leiden cohort. The Utrecht study consisted of 348 female patients with anorexia nervosa and 415 control individuals from the general population (obtained from the Immunogenetics and Transplantation Immunology Section of the Department of Immunohematology and Blood Transfusion, LUMC, Leiden, and referred to as Control Group A). Additionally, the second control group for Utrecht cases (Control Group B) consisted of 643 individuals (328 females), who were screened for not having psychiatric disorders and whole- genome genotyped (Stefansson et al., 2009). All cases and controls were of Dutch origin. Blood was collected from patients after referral to an ED treatment center (in- and outpatients, at various stages of the disease). Diagnoses, according to DSM-IV criteria, were established by experienced clinicians, with use of a semi-structured interview (Eating Disorder Examination; (Cooper& Fairburn, 1987)). Subjects for whom AN was not the
72 Chapter 4 primary diagnosis or with physical illnesses, such as diabetes mellitus, were excluded. The Leiden study consisted of 174 cases, coming from 10 specialized ED centers in the Netherlands (the GenED study), and 607 unrelated female controls from the Netherlands Twin Registry (see (Slof-Op 't Landt et al., 2011)). Diagnoses of AN were established by experienced clinicians, based on a semi-structured interview at intake and via the self-report ED examination questionnaire (EDEQ) (Fairburn& Beglin, 1994). Genotyping of Utrecht and Leiden cases and controls was done by mass spectrometry (the homogeneous MassARRAY system; Sequenom, San Diego, CA) using standard conditions, except for the Control Group B, which was genotyped on Illumina HumanHap 550k platform (Stefansson et al., 2009). The odds ratio (OR) of being a case was calculated at the level of alleles (allelic χ² test, with 1df). The allelic contrast provides more statistical power than the genotype contrasts and indicates the effect of the allele in the population (Zintzaras& Lau, 2008). Data were handled and analysed with Plink (S. Purcell et al., 2007). The study was approved by the ethical committee at UMC Utrecht (METC) and the committee for mental health institutions in the Netherlands (METiGG).
Meta-analysis Search strategy and terms A search for case-control and trio studies with genotype data for rs4680 was performed in Pubmed, Embase and ICI Web of Knowledge search engines. The following terms were used (not restricted to any fields): (COMT OR Catechol-O-methyl OR Val158Met OR Val/Met OR rs4680 OR Val108/158Met OR G1947A) AND (anorexia OR eating disorders) AND (association OR gene- association OR genetic OR polymorphism). Additionally, references in the papers of interest were searched manually. The search was last updated on 11.04.2011. To be included, a study had to report genotype frequencies of
73 Chapter 4 rs4680 in cases with AN and controls (case-control design), or allele transmission (family-based design). Where datasets were overlapping, the largest one was selected.
Data extraction In each study we extracted the data about the author, year of publication, ethnicity of participants, gender of participants, diagnostic status, sample size, genotype frequencies (case-control), the Met allele transmission (trio design). Authors were contacted if the required data were not in the article.
Statistical analysis A test for Hardy-Weinberg equilibrium was performed in each study and in the total sample (χ2 goodness-of-fit test (1df)). A meta-analysis of the binary outcome was carried out with ORs as an effect size (ES). 95% confidence intervals (CI) were estimated. Weight of each study was determined in relation to its inverse variance. Heterogeneity of ESs between studies was determined by Cochran Q-statistic (considered statistically significant for p<0.1) (Munafo& Flint, 2004) and quantified with I2 metric (I2=100%*(Q-df)/Q) (Higgins, Thompson, Deeks, & Altman, 2003). I2 ranges from 0 to 100% (from low to high heterogeneity, respectively). To determine whether the pooled ES or heterogeneity were strongly influenced by a single study we performed an influence analysis, which recalculates overall ES and I2 with each study removed per calculation. To examine the stability of the pooled ES over time the cumulative reversed analysis was performed. It recalculates overall results per each step, as the studies are added one by one in a reversed chronological order. To examine a possibility of a publication bias, a funnel plot was included and the correlation between the sample size and the ES was calculated. These should be considered with caution due to a small number of included studies (Lau, Ioannidis, Terrin, Schmid, & Olkin, 2006).
74 Chapter 4
Analyses were performed with R packages ‘catmap’ (Nicodemus, 2008) and ‘meta’ (Schwarzer, 2007). Package ‘catmap’ implements the algorithm for pooling of ESs from case-control and trio studies, as described in (Kazeem& Farrall, 2005). Genetic Power Calculator was used for calculation of power (S. Purcell, Cherny, & Sham, 2003).
Results
Association study The Utrecht cohort In the case group and both control groups subjects with more than 5% missing genotypes (genotyping was performed for multiple single nucleotide polymorphisms ( SNPs)) were removed. This resulted in exclusion of 42 out of 348 cases, 49 out of 415 controls (Control Group A) and none out of 643 controls (Control Group B). In the remaining subjects (ncases=306, ncontrols=1009), there were no missing genotypes for SNP rs4680. The OR for association of the Met allele with risk of AN in the allelic contrast was 1.14 (95%CI 0.95-1.37; P=0.14). Additionally, the dominant effect of the Met allele was tested. This produced a suggestive signal of association with OR of 1.42 (95%CI 1.02-1.96; P=0.03).
The Leiden cohort None of 174 cases and 80 of 607 controls were removed due to a threshold of 10% missed genotype calls (multiple SNPs were genotyped for these groups). Further 61 control subjects were removed due to a missed genotype call for rs4680 (88.4% genotyping success rate for rs4680), which resulted in 466 control genotypes being available for the analysis. The genotyping rate for rs4680 in the cases was 100%.
75 Chapter 4
In the Leiden cohort, the OR for association of the Met allele in the allelic contrast was 1.02 (95%CI 0.8-1.31; P=0.85). For the dominant effect of the Met allele OR was 1 (95%CI 0.63-1.58; P=1).
Genotypes were in Hardy-Weinberg equilibrium in all case and control groups (Table 2). To investigate the results from both cohorts further, we included them into a meta-analysis of all the studies testing this same association, which were available in literature. The data for two additional SNPs (rs174696 and rs165774) located in the COMT gene are available in the Supplementary Table 2 (none was significantly associated).
Meta-analysis Search results Search and inclusion of studies are shown in the flow diagram (Supplementary Figure S1). Eight studies were identified as eligible for meta- analysis (including the present genotypings). A study by Gabrovsek et al. (2004) consisted of 4 case-control cohorts and the trio part, which was not included due to an overlap with case-control groups. The samples were overlapping between Frisch et al. (2001) and Michaelovsky et al. (2005), and in Mikolajczyk et al. (2006) and Mikolajczyk, Grzywacz & Samochowiec (2010). The larger datasets were selected. Overall, there were 10 case- control sets and 1 family-based, amounting to a total of 2021 cases, 2848 controls and 89 informative trios (i.e. with heterozygous parents). Studies’ characteristics are in Table 1.
76 Chapter 4
Table 1. Characteristics of the included studies % % Case Age BMI N Study femal femal Age BMI N Study Ethnicity defini in in co type e e in AN in AN cases tion cont. cont. nt. cases cont. Karwaut DSM- 15.3 27.4 13.1 22.4 z et al., c-c British 100 100 44 38e IV (3.2)b (9.4) (2.2)c (3.8) 2001 Gabrovs 18.3 14.5 ek- Italian DSM- c-c 99 - (4.3)b, - (0.3)c, - 89 83 Florenc (Florence) IV d d e Gabrovs ek- DSM- c-c German 99 - - - - - 61 96 German IV y Gabrovs Italian DSM- c-c 99 - - - - - 51 146 ek-Milan (Milan) IV Gabrovs DSM- ek- c-c Spanish 99 - - - - - 65 93 IV Spain Dmitrza DSM- k- IV 18.5 34.9 Węglarz c-c Polish 100 100 - - 91 135 ICD- (3.2) (10) et al., 10 2005 Mikolajc DSM- zyk et IV 22.07 22.85 15.05 20.65 c-c Polish 100 100 61 105 al., ICD- (3.76 (3.95) (2.54) (1.25) 2010 10 Pinheiro European DSM- 27.1 26.3 14.7 22.1 et al., c-c 100 100 1079 677 descent IV (8.8) (8.3) (2.5)4 (1.8) 2010 Leiden DSM- 28 23.4 16.7 22.6 cohort c-c Dutch 100 100 174 466 IV (10) (12.1) (2.9) (14.1) 2011a Utrecht 14.65 DSM- 24.13 100 cohort c-c Dutch 100 33 (1.72) 306 IV (4.74) 9 2011a c Inf.
Trios Michael ovsky et DSM- 15.96 14.71 trio Israeli 100 - - - 89 al., IV (2.33) (1.71) 2005 a present genotyping; b age at onset of AN; c lifetime min BMI; d age and BMI is the average of all the cohorts present in (Gabrovsek et al., 2004); e controls were sisters of the probands; c-c, case-control design; cont., controls; BMI, body mass index; Inf. Trios, informative trios (i.e. with heterozygous parents).
77 Chapter 4
Table 2. Genotypes (counts and frequencies) and Hardy-Weinberg equilibrium in each study
cases (%) cont. (%) cases (%) cont. (%) HWE HWE Study Met Val Met allele Met/Met Val/Met Val/Val Met/Met Val/Met Val/Val cases cont. Val allele (G) allele allele (A) (A/A) (G/A) (G/G) (A/A) (G/A) (G/G) (A) (G)
Karwautz et 36 40 21 45 (51.1) 43 (48.9) 12 (27.3) 11 (25) 8 (21.1) 20 (52.6) 10 (26.3) 0.77 0.73 al., 2001 (47.4) (52.6) (47.7) Gabrovsek- 42 28 80 (44.9) 98 (55.1) 78 (47) 88 (53) 19 (21.3) 17 (20.5) 44 (53) 22 (26.5) 0.66 0.56 Florence (47.2) (31.5) Gabrovsek- 107 85 26 13 70 (57.4) 52 (42.6) 22 (36.1) 31 (32.3) 45 (46.9) 20 (20.8) 0.32 0.62 Germany (55.7) (44.3) (42.6) (21.3) Gabrovsek- 143 149 20 16 50 (49) 52 (51) 15 (29.4) 40 (27.4) 63 (43.2) 43 (29.5) 0.12 0.1 Milan (49) (51) (39.2) (31.4) Gabrovsek- 90 96 30 17 66 (50.8) 64 (49.2) 18 (27.7) 23 (24.7) 44 (47.3) 26 (28) 0.54 0.61 Spain (48.4) (51.6) (46.2) (26.2) Dmitrzak- 136 134 45 26 Węglarz et 85 (46.7) 97 (53.3) 20 (22) 25 (18.5) 86 (63.7) 24 (17.8) 0.95 0.001 (50.4) (49.6) (49.5) (28.6) al., 2005 Mikolajczyk 118 92 33 13 63 (51.6) 59 (48.4) 15 (24.6) 35 (33.3) 48 (45.7) 22 (21) 0.52 0.46 et al., 2010 (56.2) (43.8) (54.1) (21.3) Pinheiro et 665 689 286 528 265 313 188 1100 (51) 1058 (49) 176 (26) 0.49 0.05 al., 2010 (49.1) (50.9) (26.5) (48.9) (24.6) (46.2) (27.8) Leiden 533 399 85 31 150 201 (57.8) 147 (42.2) 58 (33.3) 233 (50) 83 (17.8) 0.99 0.65 cohort 2011a (57.2) (42.8) (48.9) (17.8) (32.2) Utrecht 1040 978 164 56 274 492 243 336 (54.9) 276 (45.1) 86 (28.1) 0.15 0.45 cohort 2011a (51.5) (48.5) (53.6) (18.3) (27.2) (48.8) (24.1) Total Trans. Met Untrans. Met Michaelovsk 35 (39.3) 54 (60.7) y et al., 2005 a present genotyping; c-c, case-control design; cont., controls; (Un)trans., (un)transmitted allele; HWE, P for Hardy-Weinberg equilibrium test (χ2 goodness-of-fit test, 1df).
78 Chapter 4
Test for Hardy-Weinberg equilibrium (Pearson's chi2, df=1) was significant for the control group in Dmitrzak-Weglarz et al. (2005)(P=0.001; Table 2). Exclusion of this study did not materially affect the results (Supplementary Figure S2 and S7).
Patients were diagnosed according to DSM-IV in each study (American Psychiatric Association, 2000).
Publication bias Visual inspection of the funnel plot revealed that one study (Michaelovsky et al., 2005) was located outside of the pseudo-95% confidence intervals, which indicates a possibility of a publication bias. Correlation between the weight and the effect size was non-significant (n=11, r=0.32, P=0.33)(Fig. 1).
Fig. 1 Funnel plot for 11 datasets (10 case-control, 1 family-based). Each dot represents single cohort. Location outside the delineated triangle (pseudo 95% confidence intervals) suggests a publication bias.
Heterogeneity and pooled effect size A formal test for heterogeneity of effect sizes was non-significant, with Cochrane Q statistic of 8.51 (P=0.58; I2 =0%). Therefore, meta-analysis was
79 Chapter 4 performed under the fixed-effect model (Mantel& Haenszel, 1959). This model assumes that differences in the ESs between studies are attributable to a sampling error and the true effect is homogeneous across populations. The Met allele is considered the reference allele (OR larger than 1 indicates that it is associated with increased risk of being a case). Studies were weighed using the inverse variance method. Meta-analysis of 11 cohorts (8 studies) in the allelic contrast resulted in a non-significant pooled OR of 1.03 (95%CI 0.95-1.13; P=0.42) (Fig. 2).
Fig. 2 Forest plot showing ORs in the allelic contrast (the Met allele as the risk variant). The weight of the studies is reflected by the size of squares, and whiskers represent 95% confidence intervals. The pooled OR under the fixed effect model. I2, as a measure of heterogeneity, was 0%.
The reversed cumulative analysis and the influence analysis show that the pooled results were consistent over time and were not overly impacted by any of the single datasets (Supplementary Figures S2 and S3). Since the effect size in the present association study was larger when testing for the dominant effect of the Met allele, meta-analysis was also performed under this model of genetic association (without the family-based study (Michaelovsky et al., 2005)).
80 Chapter 4
The Cochran’s Q statistic for heterogeneity was 9.21, which was non- significant (P=0.42, I2=2.2%). The fixed-effect model of meta-analysis was used. The pooled OR for 10 case-control cohorts (7 studies) in the meta- analysis of the dominant effect of the Met allele was also non-significant and equaled 1.1 (95%CI 0.95-1.27; P=0.18) (Supplementary Figure S4). The reversed cumulative and the influence plots are available in the supplementary materials, again showing that the pooled results were consisted over time and not overly impacted by any of the datasets (Supplementary Figures S5 and S6).
With the total sample size of ncases=2021 and ncontrols=2848, the Met allele frequency of 48% in the European populations and assuming the alpha level at 0.05, we were able to detect an OR of 1.13 for the risk homozygote and OR of 1.26 for the heterozygote, with 80% statistical power (for the allelic contrast). True statistical power was larger, due to the contribution of the family-based studies (89 heterozygous trios).
Discussion
Since the early days of genetic association studies in psychiatry, a functional polymorphism rs4680 of the COMT gene was considered a candidate locus for association with AN (Gorwood, Bouvard, Mouren-Siméoni, Kipman & Adès, 1998). So far, the results in the literature were not conclusive and based on relatively small sample sizes. In the present paper, in order to test the hypothesis of association, we combined genotyping and association testing of cases and controls from the Netherlands with a meta-analysis of another 6 studies (9 datasets). The results of the association testing were non-significant for the allelic contrast in the Utrecht and Leiden cohorts, but an indication of an association was observed under the dominant model of genetic effect (for the Met allele) in the Utrecht cohort. However, a meta- analysis, which in total included 2021 cases, 2848 controls and 89
81 Chapter 4 informative trios revealed an absence of association under both genetic models. There was little evidence of heterogeneity of ESs among the individual studies. Overall, the quality of included studies was good. Location of the one family-based study outside of the 95% confidence intervals in the funnel plot (Michaelovsky et al., 2005) might suggest a publication bias but it also could be reflective of a different ethnical composition of this sample. Whereas all the other studies included European subjects this one had Israeli participants. The weight of this study was 3.9% and it had little influence on the overall results. Three studies in the literature reported a significant association between rs4680 and AN. Two of them were based on the Israeli population (they used an overlapping sample) – this may explain why the results differ from the other studies, which used the European populations (Frisch et al. 2001; Michaelovsky et al. 2005). Mikolajczyk et al. (2006) reported, in a small study, a significant association, but it was accompanied by violation of the Hardy-Weinberg equilibrium. The effect size appeared to be unusually large (OR>8). This calls the reliability of this finding into question, especially since in a later study, which used an overlapping sample, the association between rs4680 and AN was non-significant (Mikolajczyk et al. 2010). Only the latter study was used in the meta-analysis. By combining multiple studies, a meta-analysis is able to estimate the effect size of interest with greater power and accuracy than any single study could do. It also enables the reader to observe the changes in effect sizes of single studies over time, drawing attention to a possible winner’s curse (overestimation of the effect size in the first study reporting a given association (Nakaoka & Inoue, 2009)).A meta-analysis can utilize unpublished studies and expose possible publication biases. Providing that sufficient amount of data is available, potential confounders and moderators of the association can be investigated. A meta-analysis is not a remedy for the methodological shortcomings of the included studies, but it provides tools for detection of such studies. Phenotypic heterogeneity (or misspecification) is a serious issue, especially in psychiatrics, where the definitions of
82 Chapter 4 phenotypes often vary across studies and the diagnoses are not unambiguous. By combining many studies, a meta-analysis can decrease the impact of mis-defined phenotypes, but, for practical reasons, this may come at a price of loosing some phenotypic specificity (e.g. analysing patients with AN in total, rather than splitting into ANR and ANBP). Meta-analyses of genetic data should be particularly vigilant for the problems of population stratification (ethnical differences between cases and controls may prompt a spurious association signal) and genetic heterogeneity (the effects of genetic variants may actually vary across populations). The present design was relatively well-powered for a single SNP analysis, although an association with a very small ES (OR<1.1) cannot be ruled out. A possibility of more complex scenarios of associations, including gene x environment and epistatic interactions, should be acknowledged. Especially, a gene x environment interaction, if not detected and accounted for, can greatly reduce the overall power of the meta-analysis. The current design did not allow for testing of such interactions. Due to limited data availability, the analysis did not include subtyping of AN phenotype (AN restrictive and AN binging/purging). Additionally, possible effect moderators, such as sex and ethnicity, were not accounted for in the analysis. However, given the facts that almost all cases were female and all but one studies (this study had weight of 3.9%; Fig. 2) included European participants, the possibility of confounding was unlikely. In conclusion, the present meta-analysis provides strong evidence that SNP rs4680 does not have a main effect on susceptibility to AN.
Acknowledgements
MKB was supported by funding from the Marie Curie Research Training Network INTACT (Individually tailored stepped care for women with eating disorders; reference number: MRTN-CT-2006-035988).
83 Chapter 4
CM was supported by the Netherlands Organization for Scientific Research NWO-ZonMw (91676125). The authors thank the Price Foundation and the Price Foundation Collaborative Group for data collection, genotyping, and data analysis. The authors are indebted to the participating families for their contribution of time and effort in support of this study. The genotypic work was supported by the Netherlands Organization of Scientific Research (MW 904-61-095, 911-03-016, 917 66344 and 911-03- 012), Leiden University Medical Centre and the Centre of Medical System Biology and Netherlands Consortium for Healthy Aging both in the framework of the Netherlands Genomics Initiative (NGI).
References
Åberg, E, Fandiño-Losada, A, Sjöholm, LK, Forsell, Y, Lavebratt, C (2011). The functional Val158Met polymorphism in catechol-O- methyltransferase (COMT) is associated with depression and motivation in men from a Swedish population-based study. J Affect Disord 129:158-166. American Psychiatric Association (2000). Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision. Washington, DC:American Psychiatric Association. Birmingham, CL, Su, J, Hlynsky, JA, Goldner, EM, Gao, M (2005). The mortality rate from anorexia nervosa. Int J Eat Disord 38:143-146. Bulik, CM, Thornton, LM, Root, TL, Pisetsky, EM, Lichtenstein, P, Pedersen, NL (2010). Understanding the relation between anorexia nervosa and bulimia nervosa in a Swedish national twin sample. Biol Psychiatry 67:71- 77. Chen, J, Lipska, BK, Halim, N, Ma, QD, Matsumoto, M, Melhem, S, Kolachana, BS, Hyde, TM, Herman, MM, Apud, J, Egan, MF, Kleinman, JE, Weinberger, DR (2004). Functional analysis of genetic variation in catechol-
84 Chapter 4
O-methyltransferase (COMT): effects on mRNA, protein, and enzyme activity in postmortem human brain. Am J Hum Genet 75:807-821. Cooper, Z, Fairburn, C (1987). The eating disorder examination: A semi-structured interview for the assessment of the specific psychopathology of eating disorders. Int J Eat Disord 6:1-8. Costas, J, Sanjuán, J, Ramos-Ríos, R, Paz, E, Agra, S, Ivorra, JL, Páramo, M, Brenlla, J, Arrojo, M (2011). Heterozygosity at catechol-O- methyltransferase Val158Met and schizophrenia: New data and meta- analysis. J Psychiatr Res 45:7-14. Dmitrzak-Węglarz, M, Słopień, A, Rybakowski, F, Czerski, PM, Hauser, J, Rajewski, A (2005). An association study of the Dopamine Transporter (DAT) gene and catechol-O-methyltransferase (COMT) gene in anorexia nervosa in the Polish population. Psychiatria i Psychoterapia 3. Fairburn, CG, Beglin, SJ (1994). Assessment of eating disorders: interview or self-report questionnaire? Int J Eat Disord 16:363-370. Frieling, H, Romer, KD, Wilhelm, J, Hillemacher, T, Kornhuber, J, de Zwaan, M, Jacoby, GE, Bleich, S (2006). Association of catecholamine-O- methyltransferase and 5-HTTLPR genotype with eating disorder-related behavior and attitudes in females with eating disorders. Psychiatr Genet 16:205-208. Frisch, A, Laufer, N, Danziger, Y, Michaelovsky, E, Leor, S, Carel, C, Stein, D, Fenig, S, Mimouni, M, Apter, A, Weizman, A (2001). Association of anorexia nervosa with the high activity allele of the COMT gene: a family- based study in Israeli patients. Mol Psychiatry 6:243-245. Gabrovsek, M, Brecelj-Anderluh, M, Bellodi, L, Cellini, E, Bella, DD, Estivill, X, Fernandez-Aranda, F, Freeman, B, Geller, F, Gratacos, M, Haigh, R, Hebebrand, J, Hinney, A, Holliday, J, Hu, X, Karwautz, A, Nacmias, B, Ribases, M, Remschmidt, H, Komel, R, Sorbi, S, Tomori, M, Treasure, J, Wagner, G, Zhao, J, Collier, DA (2004). Combined family trio and case-control analysis of the COMT Val158Met polymorphism in European patients with anorexia nervosa. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics 124B:68-72.
85 Chapter 4
Gorwood, P, Bouvard, M, Mouren-Siméoni, M C, Kipman A, Adès J (1998). Genetics and anorexia nervosa: a review of candidate genes. Psychiatr Genet 8. Higgins, JPT, Thompson, SG, Deeks, JJ, Altman, DG (2003). Measuring inconsistency in meta-analyses. BMJ 327:557-560. Karwautz, A, Rabe-Hesketh, S, Hu, X, Zhao, J, Sham, P, Collier, DA, Treasure, JL (2001). Individual-specific risk factors for anorexia nervosa: a pilot study using a discordant sister-pair design. Psychol Med 31:317-329. Kazeem, GR, Farrall, M (2005). Integrating case-control and TDT studies. Ann Hum Genet 69:329-335. Kortegaard, LS, Hoerder, K, Joergensen, J, Gillberg, C, Kyvik, KO (2001). A preliminary population-based twin study of self-reported eating disorder. Psychol Med 31:361-365. Lachman, HM, Papolos, DF, Saito, T, Yu, YM, Szumlanski, CL, Weinshilboum, RM (1996). Human catechol-O-methyltransferase pharmacogenetics: description of a functional polymorphism and its potential application to neuropsychiatric disorders. Pharmacogenetics 6:243-250. Lau, J, Ioannidis, JP, Terrin, N, Schmid, CH, Olkin, I (2006). The case of the misleading funnel plot. BMJ 333:597-600. Mantel, N, Haenszel, W (1959). Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 22:719-748. Michaelovsky, E, Frisch, A, Leor, S, Stein, D, Danziger, Y, Carel, C, Fennig, S, Mimouni, M, Klauck, SM, Benner, A, Poustka, A, Apter, A, Weizman, A (2005). Haplotype analysis of the COMT-ARVCF gene region in Israeli anorexia nervosa family trios. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics 139B:45-50. Mier, D, Kirsch, P, Meyer-Lindenberg, A (2010). Neural substrates of pleiotropic action of genetic variation in COMT: a meta-analysis. Mol Psychiatry 15:918-927. Mikolajczyk, E, Grzywacz, A, Samochowiec, J (2010). The association of catechol-O-methyltransferase genotype with the phenotype of women with eating disorders. Brain Res 1307:142-148.
86 Chapter 4
Mikolajczyk, E, Smiarowska, M, Grzywacz, A, Samochowiec, J (2006). Association of Eating Disorders with Catechol- O -Methyltransferase Gene Functional Polymorphism. Neuropsychobiology 54:82-86. Munafo, MR, Flint, J (2004). Meta-analysis of genetic association studies. Trends Genet 20:439-444. Nakaoka, H, Inoue, I (2009). Meta-analysis of genetic association studies: methodologies, between-study heterogeneity and winner's curse. J Hum Genet 54:615-623. Nicodemus, K (2008). catmap: Case-control And TDT Meta-Analysis Package. BMC Bioinformatics 9:130. Papadopoulos, FC, Ekbom, A, Brandt, L, Ekselius, L (2009). Excess mortality, causes of death and prognostic factors in anorexia nervosa. The British Journal of Psychiatry 194:10-17. Pinheiro, AP, Bulik, CM, Thornton, LM, Sullivan, PF, Root, TL, Bloss, CS, Berrettini, WH, Schork, NJ, Kaye, WH, Bergen, AW, Magistretti, P, Brandt, H, Crawford, S, Crow, S, Fichter, MM, Goldman, D, Halmi, KA, Johnson, C, Kaplan, AS, Keel, PK, Klump, KL, La Via, M, Mitchell, JE, Strober, M, Rotondo, A, Treasure, J, Woodside, DB (2010). Association study of 182 candidate genes in anorexia nervosa. Am J Med Genet B Neuropsychiatr Genet. Purcell, S, Cherny, SS, Sham, PC (2003). Genetic Power Calculator: design of linkage and association genetic mapping studies of complex traits. Bioinformatics 19:149-150. Purcell, S, Neale, B, Todd-Brown, K, Thomas, L, Ferreira, MAR, Bender, D, Maller, J, Sklar, P, de Bakker, PIW, Daly, MJ, Sham, PC (2007). PLINK: A Tool Set for Whole-Genome Association and Population-Based Linkage Analyses. The American Journal of Human Genetics 81:559-575. Schwarzer, G (2007). meta: An R Package for Meta-Analysis. R News 7:40-45. Shield, AJ, Thomae, BA, Eckloff, BW, Wieben, ED, Weinshilboum, RM (2004). Human catechol O-methyltransferase genetic variation: gene resequencing and functional characterization of variant allozymes. Mol Psychiatry 9:151-160.
87 Chapter 4
Slof-Op 't Landt, MCT, Meulenbelt, I, Bartels, M, Suchiman, E, Middeldorp, CM, Houwing-Duistermaat, JJ, van Trier, J, Onkenhout, EJ, Vink, JM, van Beijsterveldt, CEM, Brandys, MK, Sanders, N, Zipfel, S, Herzog, W, Herpertz-Dahlmann, B, Klampfl, K, Fleischhaker, C, Zeeck, A, de Zwaan, M, Herpertz, S, Ehrlich, S, van Elburg, AA, Adan, RAH, Scherag, S, Hinney, A, Hebebrand, J, Boomsma, DI, van Furth, EF, Slagboom, PE (2011). Association study in eating disorders: TPH2 associates with anorexia nervosa and self- induced vomiting. Genes Brain Behav 10:236-243. Stefansson, H, Ophoff, RA, Steinberg, S, Andreassen, OA, Cichon, S, Rujescu, D, Werge, T, Pietilainen, OPH, Mors, O, Mortensen, PB, Sigurdsson, E, Gustafsson, O, Nyegaard, M, Tuulio-Henriksson, A, Ingason, A, Hansen, T, Suvisaari, J, Lonnqvist, J, Paunio, T, Borglum, AD, Hartmann, A, Fink-Jensen, A, Nordentoft, M, Hougaard, D, Norgaard-Pedersen, B, Bottcher, Y, Olesen, J, Breuer, R, Moller, H, Giegling, I, Rasmussen, HB, Timm, S, Mattheisen, M, Bitter, I, Rethelyi, JM, Magnusdottir, BB, Sigmundsson, T, Olason, P, Masson, G, Gulcher, JR, Haraldsson, M, Fossdal, R, Thorgeirsson, TE, Thorsteinsdottir, U, Ruggeri, M, Tosato, S, Franke, B, Strengman, E, Kiemeney, LA, Melle, I, Djurovic, S, Abramova, L, Kaleda, V, Sanjuan, J, de Frutos, R, Bramon, E, Vassos, E, Fraser, G, Ettinger, U, Picchioni, M, Walker, N, Toulopoulou, T, Need, AC, Ge, D, Yoon, JL, Shianna, KV, Freimer, NB, Cantor, RM, Murray, R, Kong, A, Golimbet, V, Carracedo, A, Arango, C, Costas, J, Jonsson, EG, Terenius, L, Agartz, I, Petursson, H, Nothen, MM, Rietschel, M, Matthews, PM, Muglia, P, Peltonen, L, St Clair, D, Goldstein, DB, Stefansson, K, Collier, DA (2009). Common variants conferring risk of schizophrenia. Nature 460:744-747. Stergiakouli, E, Thapar, A (2010). Fitting the pieces together: current research on the genetic basis of attention-deficit/hyperactivity disorder (ADHD). Neuropsychiatr Dis Treat 6:551-560. Strober, M, Freeman, R, Lampert, C, Diamond, J, Kaye, W (2000). Controlled family study of anorexia nervosa and bulimia nervosa: evidence of shared liability and transmission of partial syndromes. Am J Psychiatry 157:393-401.
88 Chapter 4
Wade, TD, Bulik, CM, Neale, M, Kendler, KS (2000). Anorexia Nervosa and Major Depression: Shared Genetic and Environmental Risk Factors. Am J Psychiatry 157:469-471. Zintzaras, E, Lau, J (2008). Synthesis of genetic association studies for pertinent gene–disease associations requires appropriate methodological and statistical approaches. J Clin Epidemiol 61:634-645.
89 Chapter 5
Chapter 5
A meta-analysis of circulating BDNF concentrations in anorexia nervosa
Marek K. Brandys Martien J. H. Kas Annemarie A. van Elburg Iain C. Campbell Roger A. H. Adan
The World Journal of Biological Psychiatry 2011; 12(6):444-54.
90 Chapter 5
Abstract
Objective. Brain derived neurotrophic factor (BDNF) is involved in neuroplasticity, in the homeostatic regulation of food intake and energy expenditure. It also has a role in stress responsivity and reward processing. On the basis of its involvement in these various processes, BDNF can be hypothesized to play a role in the development and maintenance of anorexia nervosa (AN). This study meta-analytically summarizes studies which investigated serum BDNF concentrations in people currently ill with AN, in comparison to healthy controls. Methods. Seven studies measuring BDNF in serum of individuals with AN (n=155) and healthy controls (n=174) were identified and included in the meta-analysis of the mean differences between case and control groups. Results. This study confirms that AN is associated with decreased serum BDNF concentrations, in comparison to healthy controls. The combined effect size (standardized mean difference, SMD) was large (SMD= -0.96; 95% CI -1.33 to -0.59; P<0.001). Significant heterogeneity of effect sizes was identified (I2=58.3%; P<0.001), which emerged as being primarily attributable to the first published study on the investigated association. Conclusions. The present meta-analytical summary of studies measuring circulating BDNF concentrations in women with AN and healthy controls confirms that it is significantly reduced in patients. Difficulties associated with the measurement of BDNF have been identified and potential confounding factors have been discussed. Current data do not allow inferences to be made about causal links between levels of circulating BDNF and AN. However, possible explanations for the relationship between BDNF and AN have been presented.
91 Chapter 5
Introduction
Anorexia nervosa Anorexia nervosa (AN) is a debilitating disease with the highest standardized mortality ratio (SMR) of all psychiatric illnesses (estimates vary between SMR=6.2-10.5) (Birmingham et al. 2005; Papadopoulos et al. 2009). Its etiology is poorly understood. In the past decade, evidence has accrued for involvement of neurotrophic proteins, in particular brain derived neurotrophic factor (BDNF), in the etiology or maintenance of AN. Support for this has come from genetic research investigating the possibility of associations between neurotrophin genes and AN (Mercader et al. 2007; Gratacòs et al. 2007). In addition, researchers have tested the hypothesis that BDNF concentrations are altered in the blood of patients with AN (in relation to controls). Studies looking into this association were based on relatively small sample sizes and an accurate estimation of the effect size is lacking. To date, there has been no attempt to meta-analytically summarize investigations of peripheral BDNF concentrations in patients with AN, in comparison to healthy control subjects. The current paper presents such a meta-analysis together with a discussion of included studies and secondly, it examines potential mechanisms linking AN and BDNF.
BDNF: genomic and molecular characteristics The BDNF gene, which is located on chromosome 11p14.1, has a complex structure. Up to 9 promoters (Pruunsild et al. 2007; Liu et al. 2006) have been identified in the rat. Each of the promoters controls alternative transcripts which are characterized by differential expression patterns that are specific to brain regions and tissues (Timmusk et al. 1993) and which are regulated by neural activity (Koppel et al. 2009). All transcripts are translated into proBDNF (Martinowich et al. 2007), which subsequently is cleaved by proteolysis into the mature BDNF protein (Seidah et al. 1996). Both products are functional in the brain (Yang et al. 2009). Alternative cleavage results in a
92 Chapter 5 truncated BDNF isoform, for which there is no known function (Carlino et al. 2010). BDNF belongs to a family of neurotrophins and operates primarily via tyrosine kinase receptors (TrkB), whereas proBDNF binds preferentially to pan-neurotrophin receptors (p75) (Binder and Scharfman 2004). BDNF is expressed in many sites in the CNS (Hofer et al. 1990; Maisonpierre et al. 1990; Wang et al. 2010) and in peripheral tissues, such as skeletal muscles, adipose tissue and internal organs (Matthews et al. 2009).
Functional roles of BDNF BDNF is essential for proliferation, differentiation and survival of neurons during development (Hyman et al. 1991) and its involvement in molecular mechanisms underlying neural plasticity and connectivity, including activity- dependent forms of synaptic plasticity, is well documented (Waterhouse and Xu 2009). BDNF affects expression of other genes, including those in the dopaminergic (Berton et al. 2006), serotonergic (Martinowich and Lu 2007) and glutamatergic systems (Carvalho et al. 2008). It also acts as a central modulator of pain (Pezet and McMahon 2006). BDNF has been implicated in the regulation of body weight and feeding behavior in humans (Lommatzsch et al. 2005) and animals (King 2006). Large genome-wide association studies strongly implicated the BDNF gene locus in the regulation of body mass index (BMI) (Thorleifsson et al. 2009; Speliotes et al. 2010). In humans who have a functional loss of one copy of the BDNF gene, hyperphagia, obesity and hyperactivity are present (Gray et al. 2006). Mice with reduced expression of BDNF display elevated anxiety (Chan et al. 2006), increased locomotor activity and aberrant eating behavior that leads to obesity (Kernie et al. 2000; Rios et al. 2001). Several studies have investigated BDNF in the context of reward and stress reactivity, drugs, and addiction (Narita et al. 2003). BDNF is present in the ventrotegmental area (VTA) and its TrkB receptors are on dopaminergic and GABA-ergic VTA neurons (Hyman et al. 1994; Numan and Seroogy 1999). Infusion of BDNF into the VTA or nucleus accumbens (NAc) increases the rewarding effects of cocaine (Nestler and Carlezon 2006). Neurotrophins are
93 Chapter 5 critical for neuronal plasticity in response to stress and they facilitate adaptive processes (Cirulli and Alleva 2009). BDNF may also be directly involved in the physiological response to stress, as stress increases its expression in parts of the hypothalamic–pituitary–adrenal axis (HPA) (Cirulli and Alleva 2009). It has been suggested that BDNF has a role in replenishing hormonal pools by stimulating synthesis of hypothalamic neurohormones (Cirulli et al. 2010). Originally, proBDNF (precursor BDNF proteins) were thought to be inactive. However, recent evidence has shown them to be important factors that can bind to neurotrophin receptors and induce effects that are opposite to those of the mature BDNF protein (Martinowich et al. 2007). The functions of peripheral BDNF are less well characterized. Neurotrophins play a role in the integration of neural, immune and endocrine signals (Nockher and Renz 2006) and in the peripheral regulation of metabolism (Chaldakov et al. 2006).
Circulating BDNF: serum vs. plasma Peripheral BDNF concentrations are usually measured in blood samples using enzyme-linked immunosorbent assays (ELISA). These assays have good reproducibility and low inter-assay and intra-subject variability (Trajkovska et al. 2007). Either plasma, serum or whole-blood can be used. Plasma consists of blood devoid of blood cells (obtained by centrifugation of anti-coagulated blood). Serum is obtained from clotted blood. During preparation of serum, the contents of thrombocytes (platelets) are likely to be released. This is important, since peripheral BDNF is stored in platelets (Fujimura et al. 2002) and although data from serum and plasma measurements are sometimes interpreted in the same way, they are not highly correlated (Lommatzsch et al. 2005). Studies in rodents (Martin et al. 2007) and humans (Piccinni et al. 2008) show that the patterns of change in BDNF concentrations in brain plasma and serum can be different. It has been suggested that plasma concentrations, more than serum concentrations, are sensitive to environmental influences (e.g. to diet) (Piccinni et al. 2008; Brunoni et al.
94 Chapter 5
2008). Finally, serum BDNF samples are more likely to undergo changes due to the duration and temperature of storage (Trajkovska et al. 2007). These findings indicate that results from serum and plasma BDNF studies are not readily comparable.
Origins of BDNF in blood and relation to cortical levels Although the brain seems to be a major source (Rasmussen et al. 2009), blood BDNF can also arise from peripheral tissues, such as platelets and endothelial cells (Lorgis et al. 2009; Sen et al. 2008) or white and brown adipose tissue (Chaldakov et al. 2009). All three BDNF isoforms are detectable in human saliva and salivary glands (Mandel et al. 2009). Peripheral BDNF can cross the blood-brain barrier (Sartorius et al. 2009; Pan et al. 1998) and studies in rodents have demonstrated that serum BDNF reflects cortical and hippocampal levels (Karege et al. 2002; Sartorius et al. 2009). Serum BDNF and brain levels undergo similar alterations throughout the life of a rat and the correlation between them is high (r=0.81) (Karege et al. 2002). Furthermore, during a circadian cycle, serum BDNF fluctuates in a manner that is similar to cortical levels (in the adult rat)(Katoh-Semba et al. 2007). Less is known about correlations between plasma or whole-blood with cortical BDNF concentrations. Although the origins of BDNF in blood are not entirely clear, current knowledge suggests that serum BDNF is a good approximation of cortical levels.
Methods
Search strategy and study selection The literature was searched via databases Medline, Embase, ICI Web of Knowledge and by a manual search through reference lists. The search terms, which were not restricted to any fields, were as follows: (anorexia OR eating disorder*) AND (BDNF OR brain derived) AND (circulating OR serum OR plasma OR blood). It was required that an article was written in English
95 Chapter 5 and compared BDNF concentrations in serum between women with AN and healthy females as controls. Articles available online before 19.04.2010 were taken into consideration (see Fig. S1 for the search flow diagram). The following data were extracted from each study: author, year, participants’ ethnicity, gender, diagnostic status, sample size, BDNF level in cases, controls and recovered cases (means and SDs), correlation between BDNF level and BMI in patients with anorexia. The articles included in the meta-analysis together with their characteristics are listed in Table I. Authors were contacted for data not disclosed in their article and asked whether they had unpublished data of relevance to the current study (the latter was not the case).
Statistical analysis The main parameter was the standardized mean difference (SMD) in the circulating BDNF concentrations between patients with AN and healthy female controls, corrected with Hedge’s adjustment for small sample bias (Hedge’s adjusted g (Hedges 1981)). SMD was chosen rather than the mean difference as it may reduce the impact of different sampling conditions. The heterogeneity of effect sizes (ESs) between studies was estimated by Cochran Q-statistic (considered statistically significant for P<0.1) (Munafo and Flint 2004) and quantified with I2 metric (I2=(Q-df)/Q) (Higgins et al. 2003). I2 ranges from 0 to 100% (from low to high heterogeneity, respectively). The weight of each study was determined in relation to its inverse variance. To determine whether the results or heterogeneity were strongly influenced by a single study, we performed an influence analysis, which recalculates overall ES and I2 with each study removed per calculation. A cumulative meta-analysis, which recalculates overall results as the studies are added one by one, was used to investigate ES changes over time. To examine a possibility of publication bias, we have included a funnel plot and calculated the correlation between the sample size and the ES. These
96 Chapter 5 should be considered with caution due to the small number of studies that were included (Lau et al. 2006) Analyses were performed with R-package ‘meta’ (Schwarzer 2007).
Results
Included studies An internet search identified 8 studies that measured circulating BDNF concentrations in patients with AN and in control groups. 7 measured BDNF in serum and one used plasma (Mercader et al. 2007). Only the studies using serum were included in the meta-analysis. Their characteristics are presented in Table I. Samples were not overlapping between studies. In total, there were 155 cases and 174 control women. Inclusion of Mercader et al. (2007) did not materially influence the conclusions of the meta-analysis but increased heterogeneity (I2=78.6% and Q=32.69). For reference purposes, this study is also shown in Table I.
Heterogeneity The heterogeneity of ESs was quantified by Cochran’s Q and I2. For all 7 studies, I2=58.3% and Q=14.38 (P for heterogeneity <0.001) (interpreted as high heterogeneity) (Higgins et al. 2003)). Heterogeneity is predominantly attributable to the earliest study (Nakazato et al. 2003) which might be a case of ‘the winner’s curse’ (inflation of the ES in the first study in a group of studies investigating the same phenomenon). Exclusion of this study decreases heterogeneity to a non-significant level of I2=10.7% and Q=5.6 (P=0.35) (Table II). In view of the significant heterogeneity of ESs, we applied a random effects meta-analysis (DerSimonian and Laird 1986). This model assumes that there is a true variance in studies’ ESs (arising not only by chance).
Publication bias
97 Chapter 5
A visual inspection of the funnel plot suggests the presence of some bias (‘the winner’s curse’) – the earliest study (Nakazato et al. 2003) reports a larger ES (in relation to the study’s precision) than the others. A formal test for a funnel plot asymmetry was significant at t (5)=-5.1; P=0.004 (a weighted linear regression of the ES on its standard error) (Egger et al. 1997).
Figure 1. Funnel plot. Each dot represents one study. Location outside the delineated triangle (pseudo 95% confidence intervals) suggests a publication bias.
Pooled effect size The ES is the standardized mean difference (SMD) in the BDNF level between the groups, adjusted for the small sample bias (Hedge’s g (Hedges 1981)). A negative ES is indicative of a lower concentration of BDNF in the patients with AN compared to the control group. Using the random-effects model (DerSimonian and Laird 1986), the pooled SMD is -0.96 with 95% confidence intervals (CI) of (-1.33; -0.59) and P<0.001 (a large ES, according to Cohen’s guidelines; Cohen 1988). An inverse variance weighing method was used. For reference, the pooled SMD (95% CI) in the fixed effect model was -0.88 (- 1.11; -0.64). Influence analysis (Table II) shows that the earliest study (Nakazato et al. 2003) has a large impact on the pooled ES and on heterogeneity. Exclusion of this study resulted in a pooled SMD (95% CI) = -
98 Chapter 5
0.79 (-1.04; -0.54) in the random effects model and a pooled SMD (95% CI) = -0.78 (-1.02; -0.54) in the fixed effect model (P in both cases <0.001). Finally, a chronologically ordered cumulative meta-analysis demonstrated that the pooled ES tends to become weaker as the more recent studies are included, but nevertheless it remains large (Fig. 3).
Figure 2. Standardized mean difference (SMD) in BDNF serum levels between patients and controls. The weight of the study is reflected by the size of squares, and whiskers represent 95% confidence intervals. The diamond represents the pooled estimate based on the random effects model.
Figure 3. Cumulative meta-analysis. Studies are added chronologically; each row represents summary results for all studies added to this point. k - number of studies at each step. SMD - standardized mean difference in BDNF serum levels between patients and controls. The diamond represents the pooled estimate based on the random effects model.
99 Chapter 5
Table 1. Summary data from studies measuring circulating BDNF in patients with anorexia nervosa and healthy controls. Values as mean (SD). rho for Time of n ANR/ n AN HC AN HC AN BDNF HC BDNF BDFN and sample Author, Year Ethnicity AN ANP HC Age Age BMI BMI ng/ml (+/-) ng/ml (+/-) BMI in AN Medication collection Nakazato et al. 19.6 20.4 14.2 20.4 11.00- Japanese 12 7/5 21 24.9 (6.75) 61.4 (19.5) .46 present in 3 cases 2003 (5.8) (2) (0.7) (2) 12.00
Monteleone et al. 20.5 22.3 15.6 22 26.65 drug free for more Italian 22 14/8 27 45.8 (29.1) no data 8.00-9.00d 2004 (5.4) (3.4) (1.8) (1.9) (12.46) than 6 weeks drug free (80% Monteleone et al. 20 22.4 15.9 21.8 28.87 Italian 27 18/9 24 50 (27.92) .01 cases) or at least 8.00-9.00 d 2005 (5.2) (3.4) (2.4) (1.8) (16.36) for 6 weeks
Nakazato et al. 19.6 20.4 14.2 20.4 drug free at blood 11.00- Japanese 13 11/2 17 14.5 (4.4) 22.1 (8.9) .07 2006 (5.8) (2) (0.7) (1.5) drawing 12.00
25.3 24.5 14.0 20.4 drug free for at 10.00- Saito et al. 2009 Japanese 19 8/11 24 20.0 (5.1) 26.0 (3.9) .649e (7.9) (5.7) (2.1) (2.4) least 8 weeks 11.00
Nakazato et al. 28.3 26.9 15.6 22.3 drug free at blood British 29 21/8 28 11.7 (4.9) 15.1 (5.5) .05 9.00-12.00 2009 (11) (5.8) (1.6) (2.5) drawing
18.9 19 14.9 21.4 drug free for at Ehrlich et al. 2009 German 33 21/12 33 6.16 (2.88) 7.41 (3.22) .31 7.30-9.30 d (3.9) (3.1) (1.4) (2.1) least 6 weeks 15.3 Mercader et al. no no no drug free at blood Spanish 21 21c 23.8 (1.37) 57.8 (27.6) 42.6 (25.1) - 9.00-9.30 2007a data data data drawing b
a Not included in the meta-analysis - BDNF measured in plasma; b Minimum lifetime BMI; c sisters of probands as controls; d overnight fast required; e Pearson’s coefficient; AN, anorexia nervosa; ANR/ANP, AN restricting type/purging type; HC, control women; BDNF, brain-derived neurotrophic factor; rho, Spearman’s correlation coefficient
100 Chapter 5
Table II Table II. Influence analysis (random effects model) - each row shows summary results when an indicated study is omitted.
N 95% Conf. p- Omitted study Ethnicity SMD tau2 I2 studies Int. valuea
Nakazato et al. 2003 k=6 Japanese -0.79 [-1.04; -0.54] < 0.001 0.01 10.7% Monteleone et al. 2004 k=6 Italian -1.00 [-1.44; -0.56] < 0.001 0.19 65.1% Monteleone et al. 2005 k=6 Italian -0.98 [-1.42; -0.54] < 0.001 0.19 65.2% Nakazato et al. 2006 k=6 Japanese -0.96 [-1.38; -0.55] < 0.001 0.17 64.9% Saito et al. 2009 k=6 Japanese -0.91 [-1.31; -0.51] < 0.001 0.14 59.9% Nakazato et al. 2009 k=6 British -1.03 [-1.47; -0.60] < 0.001 0.18 63.0% Ehrlich et al. 2009 k=6 German -1.07 [-1.44; -0.69] < 0.001 0.10 48.8%
Pooled estimate k=7 -0.96 [-1.33; -0.59] < 0.001 0.14 58.3%
a p-value for SMD; Inverse variance method; SMD, standardized mean difference; I2, heterogeneity metric (0-100%); tau2, square-root of between-study variance
Additionally, correlations between serum BDNF concentrations and BMI in currently ill patients with AN have been combined meta-analytically. Pooling of correlation coefficients as ESs from 6 studies resulted in the pooled r=0.248 (0.069-0.412) and P=0.007 (fixed-effect model). Heterogeneity of ESs was non-significant. Details and a forest plot are shown in the Supplementary Materials (Fig. S2).
Discussion of included studies and results
The current meta-analysis summarized results from 7 studies which investigated serum BDNF concentrations from patients with AN in comparison to healthy female controls. The pooled ES in those studies confirms that serum BDNF is lower in currently ill patients than in healthy controls. Inclusion of one study (Mercader et al. 2007) which examined BDNF
101 Chapter 5 concentrations in plasma did not materially change the overall outcome but increased indicators of heterogeneity. Different concentrations of BDNF in serum are present among studies (both across the AN groups and controls). Some differences may be attributable to the technical specifications of sample storage and processing e.g. Ehrlich et al. (2009) used an ELISA assay adapted to a fluorometric technique. Other factors may also have had an effect. The mean BMI’s of cases across the studies ranged from 14.2 to 15.9 and the mean age of cases ranged from 18.9 to 28.3 suggesting differences in disease severity and duration. Whether patients were free from psychotropic or antidepressive medication at the time of sampling also varied between studies. Only one study reports exclusion of patients with psychiatric comorbidity (depression) (Saito et al. 2009). Age, depression (Lee and Kim 2009) and the use of antidepressive medication (Aydemir et al. 2005) have all been reported to affect BDNF levels. Other possible confounders include ethnicity, diet (Araya et al. 2008), season, phase of the menstrual cycle, contraceptive use, or the proportion of patients with restrictive or binge/purge subtypes of AN. Blood samples were collected at a similar time of the day (morning), but the studies differed with respect to whether an overnight fast had been required (Bus et al. 2010)(see Table I). These considerations underscore the importance of careful selection of cases and precise matching of controls, together with the need for careful reporting of those details. Limited information about control groups was available in most of the studies. Matching by age, ethnicity and perhaps diet is particularly important in studies with few participants, where random sampling errors have less chance to be cancelled out. The control groups were age-matched in three papers from Nakazato et al. (2003; 2006; 2009) and one from Saito et al. (2009). Three studies (Ehrlich et al. 2009; Monteleone et al. 2005; Monteleone et al. 2004) applied non-parametric statistical tools, whereas the remaining used parametric tests without reporting whether BDNF levels were normally distributed or if the assumption of homogeneity of variances between groups had been met. It is known that BDNF concentrations tend
102 Chapter 5 not to follow a normal distribution (Ziegenhorn et al. 2007). However, analyses performed in the current study on the raw data provided by some of the authors showed that differences arising from using parametric vs. non-parametric tests were negligible. The earliest study included in the meta-analysis was by Nakazato et al. (2003). Its ES is distinctively larger than in the others and heterogeneity indicators become statistically non-significant when the study is excluded (Table II). This observation is corroborated by visual inspection of the funnel plot (Figure 1), which suggests presence of the ‘winner’s curse’. In addition, influence analysis demonstrates that ethnicity is not an important confounder in this meta-analysis (Table II). Heterogeneity was non-significant after exclusion of Nakazato et al. (2003), even though two remaining studies had Asian participants and four used European subjects. This should be seen in the context of the small total number of included studies. Interestingly, the effect in one study that measured BDNF in plasma of patients with AN (Mercader et al. 2007) was opposite to those in the studies using serum. There, plasma BNDF concentrations were higher in the AN group than in the controls (Table I). The explanation for this finding is unclear but using plasma is not comparable to serum and it is also noted that this study used the healthy sisters of cases as controls. In three studies, changes in the BDNF concentrations were observed not only in currently affected individuals with AN but also in the recovered ones. In general, BDNF concentrations appear to increase with recovery, and in the fully recovered patients, the levels are the same or even higher than in the healthy control women (see Supplementary Materials for details). Several studies have investigated a potential link between serum BDNF levels and disease severity. Most of the data supports the view that BDNF levels are negatively correlated with the eating disorder severity and depressive comorbidity (Supplementary Materials). There are several caveats regarding studies on circulating BDNF in people with AN. Many hormones and peptides are decreased in people with AN (exceptions being cortisol, ghrelin, obestatin and growth hormone in the
103 Chapter 5 restrictive type of AN (Fichter and Pirke 1986; Estour et al. 2010)); most of these effects are probably a consequence of starvation (Fichter et al. 1986). Furthermore, specificity of serum BDNF as a potential biomarker is not high since similar alterations were found in e.g. depression (Bocchio-Chiavetto et al. 2010).
General discussion
The results of the current meta-analysis confirm that serum BDNF concentrations are reduced in women affected by AN. However, the reasons for the association remain unclear. Below, we present the most plausible explanations.
BDNF and starvation There are several possible explanations that can be used to relate alterations in serum BDNF to the AN phenotype. They can be divided into those that assume state-related changes (e.g. reaction to starvation; specific to AN) and those that postulate that lower levels are essentially a trait, for example associated with reward or stress processing and /or possibly shared with several other psychiatric disorders. In this context, it is important to recognize that state and trait conceptualizations are complementary rather than exclusive in that a trait may be present but that differences are exacerbated by state. Dietary content and caloric load are reported to have marked effects on brain BDNF expression in mice (Lee et al. 1999; Yamamoto et al. 2008; Gelegen et al. 2008). For instance, in response to a restricted feeding schedule, the C57B/L6 strain of mice show increased BDNF expression in the hypothalamus, whereas in the A/J strain, BDNF expression is reduced; it is of note that the latter strain is highly susceptible to activity-based anorexia (ABA, an animal model of AN), while the C57B/L6 strain is not (Gelegen et al. 2008). Thus, an increase in hypothalamic expression of BDNF is associated
104 Chapter 5 with resistance to ABA and a decrease with vulnerability. It is not known however, whether the lower levels of BDNF that are seen in the serum of patients with AN are due to caloric and nutritional deficiencies or (for example) are associated with the body’s reaction to stress. Another issue is that BDNF has an anorexic effect and thus its decline could be an adaptation that seeks to promote food intake. Extremely low body weight and starvation are some of the hallmarks of AN. There is a strong relationship between BMI and severity of AN symptoms and there is some evidence that clinical improvement, including an increase in BMI, is accompanied by rising BDNF concentrations. A positive correlation of BDNF with BMI in patients with AN, and a negative correlation with symptom severity (see Supplementary Materials for details) may reflect a healing process – physiological efforts to reverse the damage done by starvation. Moreover, bearing in mind BDNF’s effect on satiety (Lebrun et al. 2006), it is possible that its increase along with the increase in BMI contributes to patients’ resistance towards eating and influences adherence to therapy.
AN and depression Comorbidity between AN and depression and the presence of shared genetic and environmental risk factors (Wade et al. 2000; Silberg and Bulik 2005) provide several potential explanations for the lowered levels of BDNF seen in patients with AN. Reduced serum BDNF levels in depressed patients, lowered BDNF concentrations in post-mortem brain tissue (Karege et al. 2005), decreased BDNF levels in the CNS in animal models of depression (Angelucci et al. 2005) and restoration of BDNF levels following antidepressant drug treatment (Groves 2007) have been reported. Decreased BDNF levels have also been found in other psychiatric disorders that are comorbid with AN, such as obsessive-compulsive disorder (OCD) (Maina et al. 2010). It is arguable therefore, that alterations in BDNF in blood are probably not disorder specific but are a reflection of some events associated with psychopathology, for example, alterations in neuronal plasticity. It is of note
105 Chapter 5 that treatment with antidepressants results in normalization of BDNF levels in depressed patients and contributes to recovery or alleviation of symptoms (Başterzi et al. 2009), but it is of little or no efficacy in AN (Claudino et al. 2006). It is possible that either the effects of antidepressants on neurotrophins are diminished in AN, or a rise in neurotrophin levels is not sufficient to reduce symptoms of AN.
Hyperactivity and excessive exercise A number of studies support the idea of a compensatory and anxiolytic role of (hyper)activity and excessive exercise – behaviors frequently observed in individuals with AN (Holtkamp et al. 2004). This effect may be in part mediated by changes in BDNF expression (Maron et al. 2009). For example, in rats exposed to a stressful environment throughout life, access to the running wheel is associated with increased BDNF levels in the striatum and decreased depression-like symptoms (Marais et al. 2009). This observation is supported by the fact that exercising increases BDNF concentrations in brain, plasma and skeletal muscles of rodents (Pedersen et al. 2009; Rasmussen et al. 2009; Griesbach et al. 2009). Furthermore, in panic-disordered patients (Strohle et al. 2009) and in elderly females with remitted depression (Laske et al. 2010) acute exercise was able to transiently restore decreased serum BDNF concentrations. Therefore, it is possible that hyperactivity observed in people with AN results, in part, from its mood-regulatory effects associated with changes in BDNF concentrations.
Reward functioning, stress vulnerability and BDNF The clinical presentation of AN often involves altered reward processing, heightened anxiety and stress vulnerability (Kaye et al. 2009). Anhedonia – defined as an impaired ability to experience pleasure (Keating 2010) – is also present in recovered patients (Wagner et al. 2007), which implies that reward malfunctioning is an endophenotype of AN.
106 Chapter 5
Studies in rodents have demonstrated a pivotal role for BDNF in susceptibility to social stress paradigms, implying that some of the effects of BDNF are associated with stress and reward systems. Elevated neurotrophin levels in the VTA and NAc may lead to “hyperplasticity” of mesolimbic pathways resulting in a development and consolidation of maladaptive reactions. Modulation of stress-reactivity via mesolimbic pathways is connected to brain reward and motivation systems and associated with bidirectional projections between the VTA and the NAc, and the VTA and the PFC (Berton et al. 2006; Eisch et al. 2003). It is possible that alterations in BDNF expression, arising from interaction between genetic predisposition and adverse environmental exposures (Cirulli et al. 2009), contribute to impaired reward processing in patients (Wagner et al. 2007; Harrison et al. 2010). Bodily responses to the stress of starvation and excessive exercise, via activation of the HPA system and its effects on the VTA, stimulate otherwise anhedonic reward mechanisms, perhaps engendering a phenomenon of “addiction to starvation” (Bergh and Sodersten 1996). Hyperactivity and exercising result in a transient restoration of BDNF levels (Seifert et al. 2009) and have anxiolytic effects (Dellava et al. 2010). In parallel, excessive exercising and rejection of eating temporarily satisfies the striving for control, self-worth and high standards of performance – characteristics frequently observed in patients with AN (Sternheim et al. 2010). Stress oversensitivity results in exhaustion of psychological and physiological resources and sets up a scenario for the creation of positive feedback loops that help to maintain the illness. In depression, where the problem of stress is a core component, antidepressant medications bring improvement. In people with AN, however, inability to cope with stress may only create a milieu for the onset of other pathological mechanisms (this fact may partially explain low efficacy of antidepressants in therapy of AN (Claudino et al. 2006)). The relation between BDNF and AN fits into the diathesis-stress model (Cirulli et al. 2009), in which genetic and developmental effects of BDNF are exacerbated by adverse life events. Initially, disadvantageous pre-
107 Chapter 5 and postnatal environments promote the development of vulnerability (via maladaptations of stress and reward processing systems (Favaro et al. 2010). Subsequent life events (such as adolescent hormonal and psychological volatility, dieting) trigger the illness. The model offers a partial explanation for some of the characteristics of AN. Whether association of BDNF with AN is also mediated directly via homeostatic food regulatory circuitry is currently unclear, although there is supportive evidence from animal studies. Future studies would benefit from including self-report or test-based measures of reward, stress functioning (including early-life adverse experiences) and activity levels. Furthermore, it will be of great interest to examine other BDNF gene products, such as proBDNF, the truncated BDNF isoform as well as ratios between them and the mature protein (Carlino et al. 2010). Longitudinal investigations of BDNF changes in the course of treatment are also likely to be of interest. Larger sample sizes will be needed to elucidate the relation between BDNF concentrations with illness progression, pharmacotherapy and psychotherapy.
Conclusions
The present meta-analytical summary of studies measuring circulating BDNF concentrations in women with AN and healthy controls confirms that it is significantly reduced. It suggests that studies measuring BDNF concentrations in plasma and serum are not readily comparable. Difficulties associated with the measurement of BDNF have been identified and potential confounding factors have been discussed. Current data do not allow inferences to be made about causal links between levels of circulating BDNF and AN. However, possible explanations for the relationship between BDNF and AN have been presented, for example on the role of BDNF in mesolimbic dopamine circuits and its effects on stress responsivity and reward.
108 Chapter 5
Acknowledgements
This work was supported by funding from the Marie Curie Research Training Network INTACT (Individually tailored stepped care for women with eating disorders; reference number: MRTN-CT-2006-035988).
References
Altman SE, Shankman SA. 2009. What is the association between obsessive–compulsive disorder and eating disorders? Clin Psychol Rev 29:638-646. Angelucci F, Brene S, Mathe AA. 2005. BDNF in schizophrenia, depression and corresponding animal models. Mol Psychiatry 10:345-352. Araya AV, Orellana X, Espinoza J. 2008. Evaluation of the effect of caloric restriction on serum BDNF in overweight and obese subjects: Preliminary evidences. Endocrine 33:300-304. Aydemir O, Deveci A, Taneli F. 2005. The effect of chronic antidepressant treatment on serum brain-derived neurotrophic factor levels in depressed patients: A preliminary study. Prog Neuro-Psychopharmacol Biol Psychiatry 29:261-265. Başterzi AD, Yazici K, Aslan E, Delialioğlu N, Taşdelen B, Tot Acar Ş, et al. 2009. Effects of fluoxetine and venlafaxine on serum brain derived neurotrophic factor levels in depressed patients. Prog Neuro- Psychopharmacol Biol Psychiatry 33:281-285. Bergh C, Sodersten P. 1996. Anorexia nervosa, self-starvation and the reward of stress. Nat Med 2:21-22. Berton O, McClung CA, DiLeone RJ, Krishnan V, Renthal W, Russo SJ, et al. 2006. Essential role of BDNF in the mesolimbic dopamine pathway in social defeat stress. Science 311:864-868. Binder DK, Scharfman HE. 2004. Brain-derived neurotrophic factor. Growth Factors 22:123-131.
109 Chapter 5
Birmingham CL, Su J, Hlynsky JA, Goldner EM, Gao M. 2005. The mortality rate from anorexia nervosa. Int J Eat Disord 38:143-146. Bocchio-Chiavetto L, Bagnardi V, Zanardini R, Molteni R, Gabriela Nielsen M, Placentino A, et al. 2010. Serum and plasma BDNF levels in major depression: A replication study and meta-analyses. World J Biol Psychiatry. Brunoni AR, Lopes M, Fregni F. 2008. A systematic review and meta- analysis of clinical studies on major depression and BDNF levels: Implications for the role of neuroplasticity in depression. Int J Neuropsychopharmacol 11:1169-1180. Bus BAA, Molendijk ML, Penninx BJWH, Buitelaar JK, Kenis G, Prickaerts J, et al. 2010. Determinants of serum brain-derived neurotrophic factor. Psychoneuroendocrinology. Carlino D, Leone E, Di Cola F, Baj G, Marin R, Dinelli G, et al. 2010. Low serum truncated-BDNF isoform correlates with higher cognitive impairment in schizophrenia. J Psychiatr Res. Carvalho AL, Caldeira MV, Santos SD, Duarte CB. 2008. Role of the brain- derived neurotrophic factor at glutamatergic synapses. Br J Pharmacol 153 Suppl 1:S310-24. Chaldakov GN, Fiore M, Tonchev AB, Aloe L. 2006. Adipopharmacology, a novel drug discovery approach: A metabotrophic perspective. Letters in Drug Design & Discovery 3(7):503-505. Chaldakov GN, Tonchev AB, Aloe L. 2009. NGF and BDNF: From nerves to adipose tissue, from neurokines to metabokines. Riv Psichiatr 44(2):79-87. Chan JP, Unger TJ, Byrnes J, Rios M. 2006. Examination of behavioral deficits triggered by targeting BDNF in fetal or postnatal brains of mice. Neuroscience 142:49-58. Cirulli F, Berry A, Bonsignore LT, Capone F, D'Andrea I, Aloe L, et al. 2010. Early life influences on emotional reactivity: Evidence that social enrichment has greater effects than handling on anxiety-like behaviors, neuroendocrine responses to stress and central BDNF levels. Neurosci Biobehav Rev 34:808-820.
110 Chapter 5
Cirulli F, Alleva E. 2009. The NGF saga: From animal models of psychosocial stress to stress-related psychopathology. Front Neuroendocrinol 30:379-395. Cirulli F, Francia N, Berry A, Aloe L, Alleva E, Suomi SJ. 2009. Early life stress as a risk factor for mental health: Role of neurotrophins from rodents to non-human primates. Neuroscience & Biobehavioral Reviews 33:573-585. Claudino A, Silva de Lima M, Hay PJ, Bacaltchuk J, Schmidt U, Treasure J. 2006. In: Antidepressants for anorexia nervosa. Cohen J. 1988. Statistical power analysis for the behavioral sciences (2nd ed.). Lawrence Erlbaum Associates. Dellava JE, Thornton LM, Hamer RM, Strober M, Plotnicov K, Klump KL, et al. 2010. Childhood anxiety associated with low BMI in women with anorexia nervosa. Behav Res Ther 48. DerSimonian R, Laird N. 1986. Meta-analysis in clinical trials. Control Clin Trials 7:177-188. Egger M, Davey Smith G, Schneider M, Minder C. 1997. Bias in meta- analysis detected by a simple, graphical test. BMJ 315:629-634. Ehrlich S, Salbach-Andrae H, Eckart S, Merle JV, Burghardt R, Pfeiffer E, et al. 2009. Serum brain-derived neurotrophic factor and peripheral indicators of the serotonin system in underweight and weight-recovered adolescent girls and women with anorexia nervosa. J Psychiatry Neurosci 34:323-329. Eisch AJ, Bolanos CA, de Wit J, Simonak RD, Pudiak CM, Barrot M, et al. 2003. Brain-derived neurotrophic factor in the ventral midbrain-nucleus accumbens pathway: A role in depression. Biol Psychiatry 54:994-1005. Estour B, Germain N, Diconne E, Frere D, Cottet-Emard J, Carrot G, et al. 2010. Hormonal profile heterogeneity and short-term physical risk in restrictive anorexia nervosa. J Clin Endocrinol Metab. Favaro A, Tenconi E, Santonastaso P. 2010. The interaction between perinatal factors and childhood abuse in the risk of developing anorexia nervosa. Psychol Med 40:657-665.
111 Chapter 5
Fichter MM, Pirke KM. 1986. Effect of experimental and pathological weight loss upon the hypothalamo-pituitary-adrenal axis. Psychoneuroendocrinology 11:295-305. Fichter MM, Pirke K, Holsboer F. 1986. Weight loss causes neuroendocrine disturbances: Experimental study in healthy starving subjects. Psychiatry Res 17:61-72. Fujimura H, Altar CA, Chen R, Nakamura T, Nakahashi T, Kambayashi J, et al. 2002. Brain-derived neurotrophic factor is stored in human platelets and released by agonist stimulation. Thromb Haemost 87:728-734. Gelegen C, van den Heuvel J, Collier DA, Campbell IC, Oppelaar H, Hessel E, et al. 2008. Dopaminergic and brain-derived neurotrophic factor signalling in inbred mice exposed to a restricted feeding schedule. Genes, Brain and Behavior 7:552-559. Gratacòs M, González JR, Mercader JM, de Cid R, Urretavizcaya M, Estivill X. 2007. Brain-derived neurotrophic factor Val66Met and psychiatric disorders: Meta-analysis of case-control studies confirm association to substance-related disorders, eating disorders, and schizophrenia. Biological Psychiatry 61:911-922. Gray J, Yeo GSH, Cox JJ, Morton J, Adlam AR, Keogh JM, et al. 2006. Hyperphagia, severe obesity, impaired cognitive function, and hyperactivity associated with functional loss of one copy of the brain-derived neurotrophic factor (BDNF) gene. Diabetes 55:3366-3371. Griesbach GS, Hovda DA, Gomez-Pinilla F. 2009. Exercise-induced improvement in cognitive performance after traumatic brain injury in rats is dependent on BDNF activation. Brain Res 1288:105-115. Groves JO. 2007. Is it time to reassess the BDNF hypothesis of depression? Mol Psychiatry 12:1079-1088. Harrison A, O'Brien N, Lopez C, Treasure J. 2010. Sensitivity to reward and punishment in eating disorders. Psychiatry Res. Hedges LV. 1981. Distribution theory for glass's estimator of effect size and related estimators. Journal of Educational and Behavioral Statistics 6:107-128.
112 Chapter 5
Higgins JPT, Thompson SG, Deeks JJ, Altman DG. 2003. Measuring inconsistency in meta-analyses. BMJ 327:557-560. Hofer M, Pagliusi SR, Hohn A, Leibrock J, Barde YA. 1990. Regional distribution of brain-derived neurotrophic factor mRNA in the adult mouse brain. EMBO J 9:2459-2464. Holtkamp K, Hebebrand J, Herpertz-Dahlmann B. 2004. The contribution of anxiety and food restriction on physical activity levels in acute anorexia nervosa. Int J Eat Disord 36:163-171. Hyman C, Juhasz M, Jackson C, Wright P, Ip N, Lindsay R. 1994. Overlapping and distinct actions of the neurotrophins BDNF, NT-3, and NT- 4/5 on cultured dopaminergic and GABAergic neurons of the ventral mesencephalon. J Neurosci 14:335-347. Hyman C, Hofer M, Barde Y, Juhasz M, Yancopoulos GD, Squinto SP, et al. 1991. BDNF is a neurotrophic factor for dopaminergic neurons of the substantia nigra. Nature 350:230-232. Karege F, Schwald M, Cisse M. 2002. Postnatal developmental profile of brain-derived neurotrophic factor in rat brain and platelets. Neurosci Lett 328:261-264. Karege F, Vaudan G, Schwald M, Perroud N, La Harpe R. 2005. Neurotrophin levels in postmortem brains of suicide victims and the effects of antemortem diagnosis and psychotropic drugs. Mol Brain Res 136:29-37. Katoh-Semba R, Wakako R, Komori T, Shigemi H, Miyazaki N, Ito H, et al. 2007. Age-related changes in BDNF protein levels in human serum: Differences between autism cases and normal controls. International Journal of Developmental Neuroscience 25:367-372. Kaye WH, Fudge JL, Paulus M. 2009. New insights into symptoms and neurocircuit function of anorexia nervosa. Nat Rev Neurosci 10. Keating C. 2010. Theoretical perspective on anorexia nervosa: The conflict of reward. Neuroscience & Biobehavioral Reviews 34. Kernie SG, Liebl DJ, Parada LF. 2000. BDNF regulates eating behavior and locomotor activity in mice. EMBO J 19:1290-1300.
113 Chapter 5
King BM. 2006. The rise, fall, and resurrection of the ventromedial hypothalamus in the regulation of feeding behavior and body weight. Physiol Behav 87:221-244. Koppel I, Aid-Pavlidis T, Jaanson K, Sepp M, Pruunsild P, Palm K, et al. 2009. Tissue-specific and neural activity-regulated expression of human BDNF gene in BAC transgenic mice. BMC Neurosci 10:68. Laske C, Banschbach S, Stransky E, Bosch S, Straten G, Machann J, et al. 2010. Exercise-induced normalization of decreased BDNF serum concentration in elderly women with remitted major depression. Int J Neuropsychopharmacol 1-8. Lau J, Ioannidis JP, Terrin N, Schmid CH, Olkin I. 2006. The case of the misleading funnel plot. BMJ 333:597-600. Lebrun B, Bariohay B, Moyse E, Jean A. 2006. Brain-derived neurotrophic factor (BDNF) and food intake regulation: A minireview. Autonomic Neuroscience 126-127:30-38. Lee B, Kim Y. 2009. Reduced platelet BDNF level in patients with major depression. Prog Neuro-Psychopharmacol Biol Psychiatry 33:849-853. Lee DR, Semba R, Kondo H, Goto S, Nakano K. 1999. Decrease in the levels of NGF and BDNF in brains of mice fed a tryptophan-deficient diet. Biosci Biotechnol Biochem 63:337-340. Liu Q, Lu L, Zhu X, Gong J, Shaham Y, Uhl GR. 2006. Rodent BDNF genes, novel promoters, novel splice variants, and regulation by cocaine. Brain Res 1067:1-12. Lommatzsch M, Zingler D, Schuhbaeck K, Schloetcke K, Zingler C, Schuff-Werner P, et al. 2005. The impact of age, weight and gender on BDNF levels in human platelets and plasma. Neurobiol Aging 26:115-123. Lorgis L, Amoureux S, de Maistre E, Sicard P, Bejot Y, Zeller M, et al. 2009. Serum brain-derived neurotrophic factor and platelet activation evaluated by soluble P-selectin and soluble CD-40-ligand in patients with acute myocardial infarction. Fundam Clin Pharmacol. Maina G, Rosso G, Zanardini R, Bogetto F, Gennarelli M, Bocchio- Chiavetto L. 2010. Serum levels of brain-derived neurotrophic factor in drug-
114 Chapter 5 naïve obsessive–compulsive patients: A case–control study. J Affect Disord 122. Maisonpierre PC, Belluscio L, Friedman B, Alderson RF, Wiegand SJ, Furth ME, et al. 1990. NT-3, BDNF, and NGF in the developing rat nervous system: Parallel as well as reciprocal patterns of expression. Neuron 5:501- 509. Mandel AL, Ozdener H, Utermohlen V. 2009. Identification of pro- and mature brain-derived neurotrophic factor in human saliva. Arch Oral Biol 54:689-695. Marais L, Stein DJ, Daniels WM. 2009. Exercise increases BDNF levels in the striatum and decreases depressive-like behavior in chronically stressed rats. Metab Brain Dis. Maron E, Toru I, Maemets K, Sepp S, Vasar V, Shlik J, et al. 2009. CCK- 4-induced anxiety but not panic is associated with serum brain-derived neurotrophic factor in healthy subjects. J Psychopharmacol 23:460-464. Martin B, Pearson M, Kebejian L, Golden E, Keselman A, Bender M, et al. 2007. Sex-dependent metabolic, neuroendocrine, and cognitive responses to dietary energy restriction and excess. Endocrinology 148:4318-4333. Martinowich K, Lu B. 2007. Interaction between BDNF and serotonin: Role in mood disorders. Neuropsychopharmacology 33:73-83. Martinowich K, Manji H, Lu B. 2007. New insights into BDNF function in depression and anxiety. Nat Neurosci 10:1089-1093. Matthews VB, Astrom MB, Chan MH, Bruce CR, Krabbe KS, Prelovsek O, et al. 2009. Brain-derived neurotrophic factor is produced by skeletal muscle cells in response to contraction and enhances fat oxidation via activation of AMP-activated protein kinase. Diabetologia 52:1409-1418. Mercader JM, Ribasés M, Gratacòs M, González JR, Bayés M, Cid Rd, et al. 2007. Altered brain-derived neurotrophic factor blood levels and gene variability are associated with anorexia and bulimia. Genes, Brain and Behavior 6:706-716. Monteleone P, Fabrazzo M, Martiadis V, Serritella C, Pannuto M, Maj M. 2005. Circulating brain-derived neurotrophic factor is decreased in
115 Chapter 5 women with anorexia and bulimia nervosa but not in women with binge- eating disorder: Relationships to co-morbid depression, psychopathology and hormonal variables. Psychol Med 35:897-905. Monteleone P, Tortorella A, Martiadis V, Serritella C, Fuschino A, Maj M. 2004. Opposite changes in the serum brain-derived neurotrophic factor in anorexia nervosa and obesity. Psychosom Med 66:744-748. Munafo MR, Flint J. 2004. Meta-analysis of genetic association studies. Trends Genet 20:439-444. Nakazato M, Tchanturia K, Schmidt U, Campbell IC, Treasure J, Collier DA, et al. 2009. Brain-derived neurotrophic factor (BDNF) and set-shifting in currently ill and recovered anorexia nervosa (AN) patients. Psychol Med 39:1029-1035. Nakazato M, Hashimoto K, Shimizu E, Kumakiri C, Koizumi H, Okamura N, et al. 2003. Decreased levels of serum brain-derived neurotrophic factor in female patients with eating disorders. Biol Psychiatry 54:485-490. Nakazato M, Hashimoto K, Yoshimura K, Hashimoto T, Shimizu E, Iyo M. 2006. No change between the serum brain-derived neurotrophic factor in female patients with anorexia nervosa before and after partial weight recovery. Prog Neuro-Psychopharmacol Biol Psychiatry 30:1117-1121. Narita M, Aoki K, Takagi M, Yajima Y, Suzuki T. 2003. Implication of brain-derived neurotrophic factor in the release of dopamine and dopamine- related behaviors induced by methamphetamine. Neuroscience 119:767- 775. Nestler EJ, Carlezon J,William A. 2006. The mesolimbic dopamine reward circuit in depression. Biol Psychiatry 59:1151-1159. Nockher WA, Renz H. 2006. Neurotrophins in allergic diseases: From neuronal growth factors to intercellular signaling molecules. J Allergy Clin Immunol 117:583-589. Numan S, Seroogy KB. 1999. Expression of trkB and trkC mRNAs by adult midbrain dopamine neurons: A double-label in situ hybridization study. J Comp Neurol 403:295-308.
116 Chapter 5
Pan W, Banks WA, Fasold MB, Bluth J, Kastin AJ. 1998. Transport of brain-derived neurotrophic factor across the blood–brain barrier. Neuropharmacology 37:1553-1561. Papadopoulos FC, Ekbom A, Brandt L, Ekselius L. 2009. Excess mortality, causes of death and prognostic factors in anorexia nervosa. The British Journal of Psychiatry 194:10-17. Pedersen BK, Pedersen M, Krabbe KS, Bruunsgaard H, Matthews VB, Febbraio MA. 2009. Role of exercise-induced BDNF production in the regulation of energy homeostasis. Experimental Physiology. Pezet S, McMahon SB. 2006. Neurotrophins: Mediators and modulators of pain. Annu Rev Neurosci 29:507-538. Piccinni A, Marazziti D, Catena M, Domenici L, Del Debbio A, Bianchi C, et al. 2008. Plasma and serum brain-derived neurotrophic factor (BDNF) in depressed patients during 1 year of antidepressant treatments. J Affect Disord 105:279-283. Pruunsild P, Kazantseva A, Aid T, Palm K, Timmusk T. 2007. Dissecting the human BDNF locus: Bidirectional transcription, complex splicing, and multiple promoters. Genomics 90:397-406. Rasmussen P, Brassard P, Adser H, Pedersen MV, Leick L, Hart E, et al. 2009. Evidence for a release of BDNF from the brain during exercise. Experimental Physiology. Rios M, Fan G, Fekete C, Kelly J, Bates B, Kuehn R, et al. 2001. Conditional deletion of brain-derived neurotrophic factor in the postnatal brain leads to obesity and hyperactivity. Mol Endocrinol 15:1748-1757. Saito S, Watanabe K, Hashimoto E, Saito T. 2009. Low serum BDNF and food intake regulation: A possible new explanation of the pathophysiology of eating disorders. Prog Neuro-Psychopharmacol Biol Psychiatry 33:312-316. Sartorius A, Hellweg R, Litzke J, Vogt M, Dormann C, Vollmayr B, et al. 2009. Correlations and discrepancies between serum and brain tissue levels of neurotrophins after electroconvulsive treatment in rats. Pharmacopsychiatry 42:270-276.
117 Chapter 5
Schwarzer G. 2007. Meta: An R package for meta-analysis. R News 7:40-45. Seidah NG, Benjannet S, Pareek S, Chrétien M, Murphy RA. 1996. Cellular processing of the neurotrophin precursors of NT3 and BDNF by the mammalian proprotein convertases. FEBS Lett 379:247-250. Seifert T, Brassard P, Wissenberg M, Rasmussen P, Nordby P, Stallknecht B, et al. 2009. Endurance training enhances BDNF release from the human brain. Am J Physiol Regul Integr Comp Physiol. Sen S, Duman R, Sanacora G. 2008. Serum brain-derived neurotrophic factor, depression, and antidepressant medications: Meta- analyses and implications. Biol Psychiatry 64:527-532. Silberg JL, Bulik CM. 2005. The developmental association between eating disorders symptoms and symptoms of depression and anxiety in juvenile twin girls. J Child Psychol Psychiatry 46:1317-1326. Speliotes EK, Willer CJ, Berndt SI, Monda KL, Thorleifsson G, Jackson AU, et al. 2010. Association analyses of 249,796 individuals reveal 18 new loci associated with body mass index. Nat Genet 42:937-948. Sternheim L, Konstantellou A, Startup H, Schmidt U. 2010. What does uncertainty mean to women with anorexia nervosa? an interpretative phenomenological analysis. Eur Eat Disord Rev. Strohle A, Stoy M, Graetz B, Scheel M, Wittmann A, Gallinat J, et al. 2009. Acute exercise ameliorates reduced brain-derived neurotrophic factor in patients with panic disorder. Psychoneuroendocrinology. Thorleifsson G, Walters GB, Gudbjartsson DF, Steinthorsdottir V, Sulem P, Helgadottir A, et al. 2009. Genome-wide association yields new sequence variants at seven loci that associate with measures of obesity. Nat Genet 41:18-24. Timmusk T, Palm K, Metsis M, Reintam T, Paalme V, Saarma M, et al. 1993. Multiple promoters direct tissue-specific expression of the rat BDNF gene. Neuron 10:475-489.
118 Chapter 5
Trajkovska V, Marcussen AB, Vinberg M, Hartvig P, Aznar S, Knudsen GM. 2007. Measurements of brain-derived neurotrophic factor: Methodological aspects and demographical data. Brain Res Bull 73:143-149. Wade TD, Bulik CM, Neale M, Kendler KS. 2000. Anorexia nervosa and major depression: Shared genetic and environmental risk factors. Am J Psychiatry 157:469-471. Wagner A, Aizenstein H, Venkatraman VK, Fudge J, May JC, Mazurkewicz L, et al. 2007. Altered reward processing in women recovered from anorexia nervosa. Am J Psychiatry 164:1842-1849. Wang C, Bomberg E, Billington CJ, Levine AS, Kotz CM. 2010. Brain- derived neurotrophic factor (BDNF) in the hypothalamic ventromedial nucleus increases energy expenditure. Brain Res. Waterhouse EG, Xu B. 2009. New insights into the role of brain- derived neurotrophic factor in synaptic plasticity. Molecular and Cellular Neuroscience 42:. Yamamoto T, Hirayama A, Hosoe N, Furube M, Hirano S. 2008. Effects of soft-diet feeding on BDNF expression in hippocampus of mice. Bull Tokyo Dent Coll 49:185-190. Yang J, Siao C, Nagappan G, Marinic T, Jing D, McGrath K, et al. 2009. Neuronal release of proBDNF. Nat Neurosci 12:113-115. Ziegenhorn AA, Schulte-Herbrüggen O, Danker-Hopfe H, Malbranc M, Hartung H, Anders D, et al. 2007. Serum neurotrophins—A study on the time course and influencing factors in a large old age sample. Neurobiol Aging 28:1436-1445. Zoladz JA, Pilc A, Majerczak J, Grandys M, Zapart-Bukowska J, Duda K. 2008. Endurance training increases plasma brain-derived neurotrophic factor concentration in young healthy men. J Physiol Pharmacol 59 Suppl 7:119-132.
119 Chapter 5
Supplementary data
Supplementary Figure S1 Figure S1. Study selection diagram.
Supplementary Figure S2
Figure S2. Forest plot of correlation coefficients between BDNF level in serum and BMI, in patients with AN only. DerSimonian-Laird approach. Data for Monteleone 2004 were not available. Pearson’s correlation coefficient used for Saito 2009 and Spearman’s rank correlation coefficients in the remaining 5 studies. Z-scores are back-transformed to r-space. Inverse variance method was used.
120 Chapter 5
Supplementary section S1
BDNF and BMI in people with AN
The observation that serum BDNF levels are lower in patients currently ill with AN than in healthy controls raises the question whether these alterations are linked to BMI. All 8 studies considered a possible correlation between BMI and BDNF levels, but some are missing detailed data. Two studies found a positive correlation between BDNF levels and BMI in currently ill patients with AN(r-square .312 and .65, respectively) (Ehrlich et al., 2009; Saito et al., 2009), whereas the correlation was not significant in the healthy control women. Four other studies reported the same correlation to be significant when all subjects (cases plus healthy controls) were included (r-square from .30 to .57) (Nakazato et al., 2003) (Nakazato et al., 2009; Nakazato et al., 2006; Monteleone et al., 2005; Monteleone et al., 2004) but not only in patients with AN. This may be due to relatively small sample sizes and the narrow BMI range of patients with AN. Pooling of the correlation coefficients between serum BDNF and BMI in patients with AN resulted in the pooled r=0.248 (0.069-0.412); p-value=0.007 (fixed-effect model). Heterogeneity of effect sizes was non-significant, with Q=7.76 (5 df), I2=35.6% (0%-74.3%) and p-value for heterogeneity=0.17. Although the data on BDNF levels in relation to BMI in other populations are limited, studies most often report a lack of correlation or an
121 Chapter 5 inverse correlation. One large study of aged individuals of both genders found no correlation with serum BDNF in either depressed or non-depressed elders (Ziegenhorn et al., 2007). Another study looked at elderly subjects and found no correlation with serum concentrations (a negative trend was observed (Stanek et al., 2008). Similarly, no association with whole blood levels was found in individuals with BMI in normative ranges (Trajkovska et al., 2007). Higher serum BDNF levels in overweight compared to lean children have been found, but only after correcting for sex, race, pubertal status and platelet count (El-Gharbawy et al., 2006). In Bullo et al. (Bullo et al., 2007) serum BDNF concentrations were lower in morbidly obese women (BMI>40 kg/m2) compared to obese and overweight women, although there was no correlation with BMI. These data show that an association of peripheral BDNF with BMI varies across populations and is more often found in individuals at the extremes of BMI. It also suggests that there is an influence of moderators, such as age, gender or clinical status. Another issue is the direction of causality: does BMI affect BDNF levels, is it vice versa (and what are the plausible mediators, e.g. satiety, rewards functioning, mood alterations, activity levels), or are they both related to another variable? Profiles of appetite-related hormones (ghrelin, obestatin, peptide YY) varied between groups of women with AN restricting type, AN purging type and constitutively thin women (with the latter group being closest to controls), even though their BMI levels were in the same range (Germain et al., 2010; Germain et al., 2009; Germain et al., 2007). A similar pattern is likely for BDNF. This observation and the fact that a correlation between BMI and BDNF is most often found in extreme phenotypes rather than in the general population suggest that the link between BDNF and BMI is not directly causal.
Supplementary Table S1
122 Chapter 5
Table S2. Summary data from studies investigating BDNF in patients partially or fully recovered from AN. Values as mean (SD). AN recAN recA BDNF BDNF HC BDNF Author, n n n AN N HC ng/ml ng/ml ng/ml (+/- Year AN recAN HC BMI BMI BMI (+/-) (+/-) ) Nakaza 14.2 16.2 20.4 14.5 17.2 to et al., 13 13* 17 22.1 (8.9) (0.7) (1.7) (1.5) (4.4) (6.9) 2006
Nakaza 15.6 19.8 22.3 11.7 17.6 to et al., 29 18** 28 15.1 (5.5) (1.6) (1.1) (2.5) (4.9) (4.8) 2009
20.5 9.82 20** Ehrlich (1.3) (3.07) 14.9 21.4 6.16 7.41 et al., 33 33 weig (1.4) (2.1) (2.88) (3.22) 2009 ht 7.34 7* gain> (1.84) 10% * Followed up longitudinally (partial weight recovery); ** Independent (cross-sectional) group of recovered patients; AN, anorexia nervosa; HC, control women; BDNF, brain-derived neurotrophic factor; recAN, recovered from AN
Supplementary section S2
BDNF in recovered patients
Nakazato et al. (Nakazato et al., 2006) investigated longitudinal changes in BDNF patients with AN before and after partial weight recovery and although the change did not reach statistical significance, a significant correlation between BMI and serum BDNF concentrations in all subjects was reported. Similarly, Ehrlich et al. also found no difference in BDNF serum levels in patients with AN before and after partial weight recovery (Ehrlich et al., 2009), but reported a positive correlation between BDNF levels and BMI in all patients. In both studies, partially recovered patients had slightly higher levels of BDNF at the second assessment, but the difference was not statistically significant. These two longitudinal analyses used small subsets of the samples from the primarily cross-sectional studies (n=13 and 7,
123 Chapter 5 respectively); such small numbers might have been insufficient to determine whether the changes occur with partial weight recovery or not. Interestingly, cross-sectional comparisons of patients with acute AN with those who are fully recovered showed higher levels of the neurotrophin in the latter group (Ehrlich et al., 2009; Nakazato et al., 2009). Ehrlich et al. also included a group of recovered patients (defined as free from amenorrhea, bingeing or purging, psychotropic medication and maintaining a BMI within a normal range during the last 3 months). Serum BDNF in these 20 women was significantly higher compared to currently ill patients and slightly higher than in controls, but the latter did not reach statistical significance (Ehrlich et al., 2009). Furthermore, Nakazato et al. investigated BDNF levels in females recovered from AN (n=18) using a cross-sectional design (the criteria of recovery were similar to Ehrlich et al., but they had to be fulfilled for more than a year). The data suggested a similar trend – recovered from AN have higher serum BDNF than healthy controls, who, in turn, have higher levels than ill patients - but the differences did not reach statistical significance (Nakazato et al., 2009).
BDNF and disease severity
Ehrlich et al. reported a negative correlation with Drive for Thinness (r=-0.31) and Body Dissatisfaction scales (r=-0.45) from the Eating Disorder Inventory 2 (EDI-2; (Garner et al., 1983)) and with Global Severity Index from the Symptom Checklist (SC-90R; (Derogatis and Cleary, 1977)) (r=-0.28) in subjects with AN (Ehrlich et al., 2009). Nakazato et al. (Nakazato et al., 2009) reported a trend towards a negative correlation between the Eating Disorders Examination Questionnaire’s (EDEQ; (Fairburn and Beglin, 1994)) Eating Concern scale (r=-039), Hospital Anxiety and Depression Scale (HADS; (Zigmond and Snaith, 1983)) (r=-0.36) and serum BDNF in patients (Nakazato et al., 2009). Nakazato et al. found a negative correlation between BDNF and the Hamilton Depression Rating Scale (HDRS; (Hamilton, 1960)) in all subjects (including controls and cases of bulimia nervosa, r=-0.45) (Nakazato et al.,
124 Chapter 5
2003). On the other hand, there was no significant correlation with either EDI-2 in both studies from Monteleone et al. (Monteleone et al., 2005; Monteleone et al., 2004). In contrast to other studies, Nakazato et al. reported a positive correlation between serum BDNF and EDI-2 total score (r=0.5) and a lack of correlation with the HDRS measure (Nakazato et al., 2006). Beyond the studies included in the meta-analysis, one study by Mercader et al. shows a negative correlation between plasma BDNF levels and disease severity measured by SCL-90R’s Global Severity Index (r=-0.35) (Mercader et al., 2007)in patients with AN, and another one (Mercader et al., 2010) reports correlation between plasma BDNF and EDI scale Interoceptive Awareness (r=-0.57) and the total EDI score (r=-0.4).
References
Bullo, M., Peeraully, M.R., Trayhurn, P., Folch, J., Salas-Salvado, J., 2007. Circulating nerve growth factor levels in relation to obesity and the metabolic syndrome in women. Eur. J. Endocrinol. 157, 303-310 doi: 10.1530/EJE-06-0716. Derogatis, L.R., Cleary, P.A., 1977. Confirmation of the dimensional structure of the scl-90: A study in construct validation. J. Clin. Psychol. 33, 981-989 doi: 10.1002/1097-4679(197710)33:4<981::AID- JCLP2270330412>3.0.CO;2-0. Ehrlich, S., Salbach-Andrae, H., Eckart, S., Merle, J.V., Burghardt, R., Pfeiffer, E., Franke, L., Uebelhack, R., Lehmkuhl, U., Hellweg, R., 2009. Serum brain-derived neurotrophic factor and peripheral indicators of the serotonin system in underweight and weight-recovered adolescent girls and women with anorexia nervosa. J. Psychiatry Neurosci. 34, 323-329. El-Gharbawy, A.H., Adler-Wailes, D.C., Mirch, M.C., Theim, K.R., Ranzenhofer, L., Tanofsky-Kraff, M., Yanovski, J.A., 2006. Serum Brain- Derived Neurotrophic Factor Concentrations in Lean and Overweight Children and Adolescents. J. Clin. Endocrinol. Metab. 91, 3548-3552 doi: 10.1210/jc.2006-0658.
125 Chapter 5
Fairburn, C.G., Beglin, S.J., 1994. Assessment of eating disorders: interview or self-report questionnaire? Int. J. Eat. Disord. 16, 363-370. Garner, D.M., Olmstead, M.P., Polivy, J., 1983. Development and validation of a multidimensional eating disorder inventory for anorexia nervosa and bulimia. Int. J. Eat. Disord. 2, 15-34 doi: 10.1002/1098- 108X(198321)2:2<15::AID-EAT2260020203>3.0.CO;2-6. Germain, N., Galusca, B., Grouselle, D., Frere, D., Billard, S., Epelbaum, J., Estour, B., 2010. Ghrelin and Obestatin Circadian Levels Differentiate Bingeing-Purging from Restrictive Anorexia Nervosa. J. Clin. Endocrinol. Metab. doi: 10.1210/jc.2009-2196. Germain, N., Galusca, B., Grouselle, D., Frere, D., Tolle, V., Zizzari, P., Lang, F., Epelbaum, J., Estour, B., 2009. Ghrelin/obestatin ratio in two populations with low bodyweight: Constitutional thinness and anorexia nervosa. Psychoneuroendocrinology, 34. Germain, N., Galusca, B., Le Roux, C.W., Bossu, C., Ghatei, M.A., Lang, F., Bloom, S.R., Estour, B., 2007. Constitutional thinness and lean anorexia nervosa display opposite concentrations of peptide YY, glucagon-like peptide 1, ghrelin, and leptin. Am. J. Clin. Nutr. 85, 967-971. Hamilton, M., 1960. A rating scale for depression. J. Neurol. Neurosurg. Psychiatry. 23, 56-62. Mercader, J.M., Fernandez-Aranda, F., Gratacos, M., Aguera, Z., Forcano, L., Ribases, M., Villarejo, C., Estivill, X., 2010. Correlation of BDNF blood levels with interoceptive awareness and maturity fears in anorexia and bulimia nervosa patients. J. Neural Transm. doi: 10.1007/s00702-010-0377-8. Mercader, J.M., Fernandez-Aranda, F., Gratacos, M., Ribases, M., Badia, A., Villarejo, C., Solano, R., Gonzalez, J.R., Vallejo, J., Estivill, X., 2007. Blood levels of brain-derived neurotrophic factor correlate with several psychopathological symptoms in anorexia nervosa patients. Neuropsychobiology 56, 185-190 doi: 10.1159/000120623. Monteleone, P., Fabrazzo, M., Martiadis, V., Serritella, C., Pannuto, M., Maj, M., 2005. Circulating brain-derived neurotrophic factor is decreased in women with anorexia and bulimia nervosa but not in women with binge-
126 Chapter 5 eating disorder: relationships to co-morbid depression, psychopathology and hormonal variables. Psychol. Med. 35, 897-905. Monteleone, P., Tortorella, A., Martiadis, V., Serritella, C., Fuschino, A., Maj, M., 2004. Opposite Changes in the Serum Brain-Derived Neurotrophic Factor in Anorexia Nervosa and Obesity. Psychosom. Med. 66, 744-748 doi: 10.1097/01.psy.0000138119.12956.99. Nakazato, M., Tchanturia, K., Schmidt, U., Campbell, I.C., Treasure, J., Collier, D.A., Hashimoto, K., Iyo, M., 2009. Brain-derived neurotrophic factor (BDNF) and set-shifting in currently ill and recovered anorexia nervosa (AN) patients. Psychol. Med. 39, 1029-1035 doi: 10.1017/S0033291708004108. Nakazato, M., Hashimoto, K., Yoshimura, K., Hashimoto, T., Shimizu, E., Iyo, M., 2006. No change between the serum brain-derived neurotrophic factor in female patients with anorexia nervosa before and after partial weight recovery. Prog. Neuro-Psychopharmacol. Biol. Psychiatry 30, 1117- 1121 doi: DOI: 10.1016/j.pnpbp.2006.04.017. Nakazato, M., Hashimoto, K., Shimizu, E., Kumakiri, C., Koizumi, H., Okamura, N., Mitsumori, M., Komatsu, N., Iyo, M., 2003. Decreased levels of serum brain-derived neurotrophic factor in female patients with eating disorders. Biol. Psychiatry 54, 485-490 doi: DOI: 10.1016/S0006- 3223(02)01746-8. Saito, S., Watanabe, K., Hashimoto, E., Saito, T., 2009. Low serum BDNF and food intake regulation: A possible new explanation of the pathophysiology of eating disorders. Prog. Neuro-Psychopharmacol. Biol. Psychiatry 33, 312-316 doi: DOI: 10.1016/j.pnpbp.2008.12.009. Stanek, K., Gunstad, J., Leahey, T., Glickman, E., Alexander, T., Spitznagel, M., Juvancic-Heltzel, J., Murray, L., 2008. Serum brain-derived neurotrophic factor is associated with reduced appetite in healthy older adults. The Journal of Nutrition, Health and Aging 12, 183-185. Trajkovska, V., Marcussen, A.B., Vinberg, M., Hartvig, P., Aznar, S., Knudsen, G.M., 2007. Measurements of brain-derived neurotrophic factor: Methodological aspects and demographical data. Brain Res. Bull. 73, 143-149 doi: DOI: 10.1016/j.brainresbull.2007.03.009.
127 Chapter 5
Ziegenhorn, A.A., Schulte-Herbrüggen, O., Danker-Hopfe, H., Malbranc, M., Hartung, H., Anders, D., Lang, U.E., Steinhagen-Thiessen, E., Schaub, R.T., Hellweg, R., 2007. Serum neurotrophins—A study on the time course and influencing factors in a large old age sample. Neurobiol. Aging 28, 1436-1445 doi: DOI: 10.1016/j.neurobiolaging.2006.06.011. Zigmond, A.S., Snaith, R.P., 1983. The hospital anxiety and depression scale. Acta Psychiatr. Scand. 67, 361-370.
128 Chapter 6
Chapter 6
The Val66Met polymorphism of the BDNF gene in anorexia nervosa: new data and a meta-analysis
Marek K. Brandys Martien J. H. Kas Annemarie A. van Elburg Roel Ophoff Margarita C.T. Slof-Op ’t Landt Christel M. Middeldorp Dorret I. Boomsma Eric F. van Furth Eline Slagboom Roger A. H. Adan
The World Journal of Biological Psychiatry 2013; 14(6):441-51.
129 Chapter 6
Abstract
Objectives. The Val66Met polymorphism (rs6265) of the BDNF gene is a non- synonymous polymorphism, previously associated with anorexia nervosa (AN). Methods. We genotyped rs6265 in 235 patients with AN and 643 controls. Furthermore, we performed a systematic review of all case-control and family-based studies testing this SNP in AN, and combined the results in a meta-analysis. Results. The results of the case-control study were non- significant. For the meta-analysis, 9 studies were identified (ncases=2,767; ncontrols=3,322, ntrios=53) and included. Primarily, the analyses indicated an association with OR of 1.11 (P=0.024) in the allelic contrast, and OR of 1.14 (P=0.025) for the dominant effect of the Met allele. However, additional analyses revealed that the first published study (from those included in the meta-analysis) overly influenced the pooled effect size (possibly due to a phenomenon known as a winner’s curse). When this case-control study was replaced by a trio study (ntrios=293) performed on a largely overlapping sample, the effect size became smaller and non-significant, both for the allelic contrast (OR=1.07, P=0.156) and the dominant effect (OR=1.07, P=0.319). The quality of included studies was good and there was no significant heterogeneity across the effect sizes. Conclusions. Our analyses indicate that the BDNF Val66Met variant is not associated with AN at detectable levels.
130 Chapter 6
Introduction
Anorexia nervosa (AN) is a chronic and potentially lethal disorder. It is known for having the highest standardized mortality ratio of all psychiatric illnesses (mortality rate is 6-10 times higher than in a reference population (Birmingham et al. 2005; Papadopoulos et al. 2009)). Despite this seriousness, the etiology of AN remains unclear. Twin and adoption studies estimate that 46 to 78% of variance in AN is attributable to genetic factors (Kortegaard et al. 2001; Wade et al. 2000; Bulik et al. 2010) and family studies have determined a 10-fold increase in lifetime risk of developing AN for a first-degree female relative of a proband with AN (compared to relatives of unaffected individuals) (Strober et al. 2000). The main focus of genetic association studies in AN has been on candidate genes from neuropeptide and neurotransmitter pathways. The yield of those studies is very limited as many of the initially positive findings failed to replicate (Pinheiro et al. 2010). One way to deal with inconsistency of results is to perform a meta-analysis in which studies that test the same hypothesis are combined, thereby providing insight into potential sources of heterogeneity between them. The statistical power to detect an association is increased and conclusions are more solid, as compared to individual studies. A gene that has been recurrently proposed as AN susceptibility locus is the brain-derived neurotrophic factor (BDNF) gene. BDNF is crucial for proliferation, differentiation and survival of neurons during development (Hyman et al. 1991), for neuronal plasticity and connectivity in adults (including activity-dependent forms of synaptic plasticity) (Waterhouse and Xu 2009), and it affects expression of many other genes (Berton et al. 2006). BDNF is involved in energy balance and food intake regulation (Lommatzsch et al. 2005), in peripheral regulation of metabolism (Pedersen et al. 2009), and it plays a substantial role in reward and stress processing (Narita et al. 2003; Cirulli and Alleva 2009). Patients with AN have decreased levels of
131 Chapter 6
BDNF in serum (Brandys et al. 2011). All these findings make the BDNF gene an interesting candidate for studies of AN. The Val66Met (rs6265) is a non-synonymous polymorphism leading to a valine to methionine substitution in the proBDNF product (precursor of the mature BDNF protein). At the molecular level, the Met allele results in lowered intracellular trafficking and activity-dependent secretion of the BDNF protein (Egan et al. 2003; Chiaruttini et al. 2009), likely without changing its constitutive secretion (Chen et al. 2006). The Val66Met has frequently been studied in the context of (neurodevelopmental) psychiatric disorders and other behavioral traits. This polymorphism has been implicated in structural variation in human brain at the level of prefrontal cortex (PFC) and hippocampus (with the Met allele carriers having lower volumes (Szeszko et al. 2005; Pezawas et al. 2004)). A meta-analysis of the variant’s effects in major depressive disorder showed the Met allele to be associated with the condition (only in males, both in Asians and Europeans) (Verhagen et al. 2008). A meta-analysis of studies in ADHD refuted a putative involvement of the variant in the pathogenesis of that disorder (Sánchez-Mora et al. 2009). In schizophrenia, one meta-analysis rejected the association (Kawashima et al. 2009), whereas another one confirmed it (Gratacòs et al. 2007). Furthermore, a meta-analysis performed by Frustaci et al. revealed the association of the Met allele with personality traits related to anxiety (the Met allele carriers displayed lower neuroticism than non-carriers) (Frustaci et al. 2008). On the other hand, a recent, large study found that Met allele homozygotes score higher on harm avoidance (a phenotype strongly correlated with neuroticism) (Montag et al. 2010). Genetic variants associated with body mass index (BMI) are of particular interest for the AN phenotype, since extremely low body weight is the primary symptom of this illness (Brandys et al. 2009). Gunstad et al. (Gunstad et al. 2006) reported that subjects from a healthy population who carry the Val allele had higher BMI than Met/Met homozygotes. Another study, however, found that the Met allele in women was associated with obesity (Beckers et al. 2008). Contrary to Beckers et al. and in line with
132 Chapter 6
Gunstad et al., a large study showed that the Met allele was associated with lower BMI in the general population (Shugart et al. 2009). Finally, a genome- wide study of over 30,000 subjects also found the Met allele to be associated with decreased BMI in the general population (Thorleifsson et al. 2009). Despite a very large number of published studies, there remains uncertainty regarding the strength and nature of any of these associations in relation to AN. In 2007, a meta-analysis of case-control studies investigating the Val66Met polymorphism in four psychiatric diseases – substance-related disorders (SRD), EDs (AN and bulimia nervosa taken together), schizophrenia and mood disorders – implied that there is an association between the first three and rs6265 (but not with mood disorders) (Gratacòs et al. 2007). The direction of the association was opposite for EDs and substance related disorders. Whereas the Met allele increased susceptibility to ED (the fixed- effect pooled OR was 1.36 for a dominant model of genetic effect), it was also found to have a protective effect against SRD. With regards to EDs, the meta-analysis of Gratacos et al. (2007) was based on 5 datasets, with a total of 1733 cases and 1811 controls. A limitation of this study was that it did not include family-based association studies and that it considered all EDs together, whereas heterogeneity of this category is well-known (Wonderlich et al. 2007). Furthermore, since the time that the study was published, a substantial amount of data on the association between rs6265 and AN have become available. Therefore, we decided to study a more homogeneous phenotype within eating disorders – which, in our case, was AN – and include all the available data. Subsequently, a gene-association study on a sample of females with AN and healthy controls was performed and the results were combined with a meta-analysis of case-control and family based studies on the Val66Met polymorphism in subjects with AN (in total, combining 9 datasets).
133 Chapter 6
Subjects and Methods
Association study
SNP rs6265 (Val66Met) of the BDNF gene was genotyped in a sample of female cases with AN (n=235), all with ascertained Dutch descent (patients are asked whether all of their grandparents were of Dutch origin). Subjects were recruited for the study after referral to an ED treatment center (in- and outpatients, at various stages of the disease). Diagnoses were established by experienced clinicians according to DSM-IV criteria, using a semi-structured interview (Eating Disorder Examination; (Cooper and Fairburn 1987)). Cases for whom AN was not the primary diagnosis or with physical illnesses such as diabetes mellitus were excluded. 81 cases were overlapping with the sample used in de Krom et al. (de Krom et al. 2005) and they were excluded. Genotyping was performed by mass spectrometry (the homogeneous MassARRAY system; Sequenom, San Diego, CA) using standard conditions. The control group consisted of 643 Dutch individuals (328 females) screened against psychiatric disorders (Stefansson et al. 2009). Healthy controls were genotyped on the Illumina HumanHap 550k platform. Data were handled and analyzed with Plink (Purcell et al. 2007). The study has been approved by the Medical Ethical Committee at UMC Utrecht, The Netherlands.
Meta-analysis
Search strategy and terms We searched for case-control and trio studies through Pubmed, Embase and ICI Web of Knowledge search engines, using the following terms (not restricted to any fields): (bdnf OR brain derived OR Val66Met OR Val/Met OR rs6265 OR 196G/A) AND (anorexia OR eating disorders) AND (association OR gene-association OR polymorphism). Additionally, the HuGE Navigator database was checked and a manual search through references in identified
134 Chapter 6 articles performed. The search was last updated on 21st of October 2010. To be included, a study had to report genotype frequencies of rs6265 in cases with AN - be it the restricting or binging/purging type - and healthy controls (case-control design), or allele transmission (trio design). In case of overlapping datasets the larger one was selected. The meta-analysis was performed in compliance with the PRISMA statement (Moher et al. 2009).
Data extraction The following data were extracted from each study: author, year of publication, ethnicity of participants, gender of participants, diagnostic status, sample size, genotype frequencies (case-control), the Met allele transmission (trio design). Authors were contacted if the required data were not in the article.
Statistical analysis In the primary analyses, the odds ratios (ORs) were calculated at the level of alleles. Allele contrasts provide more statistical power than genotype contrasts and indicate the effect of the allele in the population (Zintzaras and Lau 2008). Since the dominant effect of the Met allele in EDs has also been suggested in the literature (Gratacòs et al. 2007), we tested for it in the additional analysis (discarding the trio studies). Hardy-Weinberg equilibrium was tested in each study and in the total sample with a χ2 goodness-of-fit test (1df). A meta-analysis of the binary outcome was performed with ORs as an effect size (ES). The Met allele was considered the risk allele. 95% confidence intervals (CI) were estimated where appropriate. Weight of each study was determined in relation to its inverse variance. Three rounds of analyses were performed. First, using the allelic contrast with the complete dataset of 9 studies (8 case-control, 1 trio); second, testing the dominant effect of the Met allele in 8 case-control studies (the dominant effect in EDs was suggested in the literature (Gratacòs et al. 2007)); third, analysis in the allelic contrast with replacement of one case-
135 Chapter 6 control study (Ribases et al. 2004a) by the trio study on partially overlapping cases (Ribases et al. 2004b). Heterogeneity of ESs between studies was estimated by Cochran Q-statistic (considered statistically significant for p<0.1) (Munafo and Flint 2004) and quantified with I2 metric (I2=100%*(Q-df)/Q) (Higgins et al. 2003). I2 ranges from 0 to 100% (from low to high heterogeneity, respectively). To determine whether the pooled ES or heterogeneity were strongly influenced by a single study we performed an influence analysis, which recalculates overall ES and I2 with each study removed per calculation. The ES in the first published report on a given association is often larger than in the later studies of the same hypothesis (Nakaoka and Inoue 2009). A reversed cumulative meta-analysis recalculates overall results as the studies are added one by one, in a reversed chronological order. It was used to investigate changes of the pooled ES, as particular ESs are reported over time. To examine the possibility of a publication bias we have included a funnel plot and calculated the correlation between the sample size and the ES. These should be considered with caution due to a small number of included studies (Lau et al. 2006). Analyses were performed with R packages ‘catmap’ (Nicodemus 2008) and ‘meta’ (Schwarzer 2007). Package ‘catmap’ implements the algorithm for pooling of ESs from case-control and trio studies, as described in (Kazeem and Farrall 2005). Genetic Power Calculator was used for calculation of power (Purcell et al. 2003).
Results
Association study
235 female cases (118 AN restrictive type; 117 AN purging type) were successfully genotyped (100% call rate). Healthy controls (n=643), genotyped
136 Chapter 6 on the Illumina HumanHap 550k platform, also had a 100% call rate for this SNP. Case and control genotypes were in Hardy-Weinberg equilibrium (Table 2). In this sample, SNP rs6265 was not associated with AN under any model of genetic effect (OR=1.058 in the allelic contrast, P=0.67, 1df).
Meta-analysis
Search results The flow diagram of the search is available in the supplementary materials (Supplementary Figure 1). The search identified 8 eligible studies (7 case-control and 1 family- based). Additionally, the results of the current genotyping were added. One sample set was studied both in a case-control and a trio design (88% overlapping cases between (Ribases et al. 2004b) and (Ribases et al. 2004a)). Only the data from the case-control approach were used in the primary analysis (due to its larger sample size). In the second round of analysis, the trio study was included instead of the case-control one. In all studies, patients were diagnosed according to DSM-IV (American Psychiatric Association 2000). The total sample size was 2,767 cases, 3,322 controls and 53 informative (with heterozygous parents) trios in the first analysis and 2,014 cases, 2,812 controls and 346 informative trios in the second analysis. All case and control groups were in Hardy-Weinberg equilibrium. Details of included studies are available in Table 1 and 2.
137 Chapter 6
TABLE I. Characteristics of the studies included in the meta-analysis. Case % % Age Age N Study BMI in BMI in Study Ethnicity definiti female female in in N cases co type AN cont. on cases cont. AN cont. nt.
Ribases et al. European, 23.6 14.2 51 c-c DSM-IV 96 95.5 753 2004a multicenter (8.9) (2.3)d 0 Koizumi et 25 27 22 c-c Japanese DSM-IV 100 100 72 al. 2004 (6)c (6) 2 26.7 Friedel et al. 24.7 17.2 21.9 c-c German DSM-IV 93.1 51 (11.5 118 96 2005 (2.6) (3.7) (1.1) ) de Krom et Dutch 26.7 15.9 58 c-c DSM-IV 100 ~50 195 al. 2005 Caucasian (2.7) (2.4) 0 Rybakowski 17.6 20.7 15.4 20.1 c-c Polish DSM-IV 100 100 144 86 et al. 2007 (2.5) (1.2) (2.2) (2.3) Pinheiro et European 27.1 26.3 14.7 22.1 67 c-c DSM-IV 100 100 1079 al. 2010 descent (8.8) (8.3) (2.5)d (1.8) 7 Slof-Op 't Dutch 28 16.7 50 Landt et al. c-c DSM-IV 100 100 171 Caucasian (10) (2.9) 8 2010 Brandys et Dutch 22.2 16.2 64 c-c DSM-IV 100 51 235 al. 2011a Caucasian (4.9) (1.8) 3 Inf. Trios
Dardennes et trio French DSM-IV 100 - N/A - 53 al. 2007 Ribases et al. European, 20.8 13.66 trio DSM-IV 97.3c - - 293 2004bb multicenter (6.7) (2.1)d a present genotyping; b study used in the second round of analyses; c for all cases with eating disorders; d lifetime min BMI; c-c – case- control design; cont. – controls; Inf. Trios – informative trios (i.e. with heterozygous parents).
138 Chapter 6
TABLE II. Counts and frequencies of genotypes and Hardy-Weinberg equilibrium in the included studies.
cases (%) cont. (%) cases (%) cont. (%) HWE HWE Study Met allele Val allele Met allele Val allele Met/Me Val/Met Val/Val Met/Met Val/Met Val/Val cases cont. (A) (G) (A) (G) t (A/A) (G/A) (G/G) (A/A) (G/A) (G/G)
Ribases et al. 175 845 268 457 145 350 324 (21.5) 1182 (78.5) 28 (3.7) 15 (2.9) 0.14 0.99 2004a (17.2) (82.8) (35.6) (60.7) (28.4) (68.6) Koizumi et 184 260 10 62 (43.1) 82 (56.9) 42 (58.3) 20 (27.8) 42 (18.9) 100 (45) 80 (36) 0.11 0.28 al. 2004 (41.4) (58.6) (13.9) Friedel et al. 157 62 42 (17.8) 194 (82.2) 35 (18.2) 5 (4.2) 32 (27.1) 81 (68.6) 1 (1) 33 (34.4) 0.48 0.13 2005 (81.8) (64.6) de Krom et 226 934 125 377 82 (21) 308 (79) 12 (6.2) 58 (29.7) 23 (4) 180 (31) 0.15 0.79 al. 2005 (19.5) (80.5) (64.1) (65) Rybakowski 147 64 50 (17.4) 238 (82.6) 25 (14.5) 3 (2.1) 44 (30.6) 97 (67.4) 3 (3.5) 19 (22.1) 0.44 0.30 et al. 2007 (85.5) (74.4) Pinheiro et 269 1085 361 676 217 434 445 (20.6) 1713 (79.4) 42 (3.9) 26 (3.8) 0.47 0.86 al. 2010 (19.9) (80.1) (33.5) (62.7) (32.1) (64.1) Slof-Op 't 192 824 114 154 335 Landt et al. 65 (19) 277 (81) 8 (4.7) 49 (28.7) 19 (3.7) 0.37 0.80 (18.9) (81.1) (66.7) (30.3) (65.9) 2010 Brandys et 259 1027 147 201 413 99 (21.1) 371 (78.9) 11 (4.7) 77 (32.8) 29 (4.5) 0.82 0.47 al. 2011a (20.1) (79.9) (62.6) (31.3) (64.2) 1365 5279 119 931 1717 1049 2115 Total 1169 (21.1) 4365 (78.9) 158 (4.8) 0.61 0.06 (20.5) (79.5) (4.3) (33.6) (62.1) (31.6) (63.7) Trans. Met Untran. Met Dardennes et 28 (52.8) 25 (47.2) al. 2007 Ribases et al. 154 (52.6) 139 (47.4) 2004bb a present genotyping; b study included in the second analyses; c-c – case-control design; cont. – controls; (Un)trans. – (un)transmitted allele; HWE – P for Hardy- Weinberg equilibrium test (χ2 goodness-of-fit test, 1df).
139 Chapter 6
Heterogeneity The hypothesis of no heterogeneity between ESs was not rejected (Cochrane Q-statistic=4.31, I2=0%, P=0.83; for all 9 studies). Likewise, there was no significant heterogeneity in the analysis of the dominant effect of the Met allele in 8 case-control studies (Q=7.2, I2=2.8%, p=0.408), nor in the second round of analysis (in which the case-control study (Ribases et al. 2004a) was replaced by the family-based study (Ribases et al. 2004b)). Therefore, the fixed-effect model of meta-analysis was applied (Mantel and Haenszel 1959). The fixed-effect model assumes that differences in the ESs between studies are attributable to a sampling error and the true effect is homogeneous across populations.
Publication bias Visual inspection of the funnel plot does not suggest presence of a publication bias, which was confirmed by a non-significant result of the linear regression test of the funnel plot asymmetry (P=0.725, for the case-control studies only). A correlation between the sample size and the effect size was non-significant (r=0.17, P=0.688 for cases and controls; r=0.26, P=0.537 for cases only).
Figure 1. Funnel plot for 9 studies (8 case-control and 1 family-based). Each dot represents one study. Location outside the delineated triangle (pseudo 95% confidence intervals) suggests a publication bias.
140 Chapter 6
Pooled effect size The OR larger than 1 indicates that the Met allele is associated with increased risk of being a case. The inverse variance weighing method and the fixed-effect model of meta-analysis were used in all analyses.
First analysis: the allelic contrast and the dominant effect In the analysis of 1 trio and 8 case-control studies, the pooled OR was 1.11 (95%CI; 1.014-1.223; P=0.024).
Figure 2. Forest plot presenting ORs for anorexia nervosa in the allelic contrast (the Met allele as the risk variant). Dardennes et al. 2007 is a family-based study, the remaining ones have a case-control design. The weight of each study is reflected by the size of squares, and whiskers represent 95% confidence intervals. The pooled OR is based on the fixed effect model. I2, as a measure of heterogeneity, equals 0%.
To further investigate the nature of the association, the influence and the reversed cumulative meta-analyses were performed on the complete set of 9 studies (using the allelic count contrast in the case-control studies).
141 Chapter 6
Figure 3. Cumulative meta-analysis. Studies are added in a reversed chronological order; each row represents the pooled OR for all studies added up to this point (based on the fixed-effect model and for the allelic contrast). The whiskers represent (cumulative) 95% confidence intervals. I2, as a measure of heterogeneity, was 0% at every step.
In the reversed cumulative meta-analysis, studies are added from the most recent to the earliest one, and the pooled ES is recalculated per each iteration. It revealed that the association remains non-significant until the earliest study (Ribases et al. 2004a) is added. Also the influence analysis, which shows the overall results with one study removed per each calculation, confirmed this (Supplementary Table S1). Both analyses show that the overall association between rs6265 and AN is attributable predominantly to the first study by Ribases et al. (2004a). The additional analysis of the dominant effect of the Met allele (Met/Met + Val/Met genotypes vs. Val/Val), in 8 case-control studies, resulted in the pooled OR of 1.138 (95%CI; 1.017-1.275; P=0.025) (Supplementary Figure S2). There, the first study was also largely responsible for driving the pooled association signal (Supplementary Figure S3; Supplementary Table S2).
142 Chapter 6
Second analysis: the allelic contrast and the dominant effect In view of the results from the influence and cumulative analyses we decided to run the meta-analysis again, this time replacing the case-control study from Ribases et al. (2004a) with a trio study by Ribases et al. (2004b) (cases between those studies are overlapping, therefore only one can be used at a time). This step was motivated by reasoning that a family-based study provides a better protection from effects of population stratification than a case-control design (at a slight loss of power, however). In this analysis, a combined ES for 9 studies was lower and it did not reach statistical significance, with the pooled OR of 1.071 (P=0.156; 95%CI; 0.974- 1.179). Heterogeneity indicators were non-significant and lower than in the first analysis – Cochrane’s Q was 1.809 (I2=0%; P=0.986).
Figure 4. Forest plot presenting ORs for anorexia nervosa in the allelic contrast (the Met allele as the risk variant). Here, the case-control study from Ribases et al. 2004a has been replaced with the trio study performed on a partially overlapping sample. Ribases et al. 2004b and Dardennes et al. 2007 are family-based, the remaining studies have a case-control design. The weight of each study is reflected by the size of squares, and whiskers represent 95% confidence intervals. The pooled OR is based on the fixed effect model. I2, as a measure of heterogeneity, equals 0%.
143 Chapter 6
Figure 5. Cumulative meta-analysis. Studies are added in a reversed chronological order; each row represents the pooled OR for all studies added to this point (based on the fixed-effect model and for the allelic contrast). ). Here, the case-control study from Ribases et al. 2004a has been replaced with the trio study performed on a partially overlapping sample. Ribases et al. 2004b and Dardennes et al. 2007 are family-based, the remaining studies have a case-control design. The whiskers represent (cumulative) 95% confidence intervals. I2, as a measure of heterogeneity, was 0% at every step.
Consistently, testing of the dominant effect without the case-control study from Ribases et al. (2004a) (7 studies included) resulted in a non- significant pooled OR of 1.068 (P=0.319; 95%CI; 0.9387; 1.2141). There was no significant heterogeneity (Q=2.98, I2=0%, P=0.811) (Supplementary Figure 4 and 5; Supplementary Table 3).
Given the total number of cases and controls (n=2,676 and n=3,322, respectively), a frequency of the Met allele in the European populations of 20%, and setting the alpha at 0.05, there was 80% power to detect an association with OR of 1.135 for the heterozygote and OR of 1.27 for the
144 Chapter 6
Met/Met homozygote (in the allelic contrast). For the dominant effect of the Met allele, with the same assumptions, there was 80% power for OR of 1.16 (for the Met allele carriers). Exclusion of the first published study on rs6265 in AN decreased power to 70%, under both scenarios. The real statistical power was slightly larger, since these calculations do not take into account the contribution from the family-based studies.
Discussion
The present study investigated a possible association between rs6265 polymorphism of the BDNF gene and AN. A meta-analytical framework was used to combine results from case-control and trio studies, with addition of new data from genotyping of the Utrecht cohort of patients with AN and healthy controls. This meta-analysis included the largest number of cases tested for association with a single SNP in AN up to date. Primary results showed that the association in question has an OR of 1.11 (a two-sided P of 0.024, in the allelic contrast). All ESs, except for one, were in the same direction (with the Met allele increasing the risk) and there was no significant heterogeneity among them. The OR became slightly more pronounced when the dominant effect of the Met allele was tested (without the trio study (Dardennes et al. 2007), in that case). Nevertheless, the influential and cumulative analyses revealed that the pooled ES was strongly influenced by the earliest study (Ribases et al. 2004a). With the first study removed, the pooled OR became closer to unity and non-significant (in accordance with the so-called ‘winners curse’, i.e. inflation of the ES in the first study in a group of studies investigating the same phenomenon (Nakaoka and Inoue 2009)). These observations suggest that the ES reported in the multicenter case-control study from Ribases et al. (2004a) might have been overestimated. By replacing this study with the trio study (Ribases et al. 2004b) of a largely overlapping set of subjects, the possibility of a cryptic population stratification was reduced. This step
145 Chapter 6 resulted in a shrinkage of the overall ES to a non-significant level and showed that the control group may be responsible for a slight overestimation of the ES in the case-control study (Ribases et al. 2004a) (regardless the fact that the controls were matched by ethnicity and sex). Frequencies of rs6265 polymorphism are very variable across populations (Petryshen et al. 2009) and a possibility of undetected population stratification is high. Similarly, when the meta-analysis was performed for the dominant effect of the Met allele, but with exclusion of the case-control study from Ribases et al. (2004a) (thus on 7 case-control studies only), the association did not reach statistical significance. Very little heterogeneity in all scenarios of analysis suggests that the quality of evidence was good. The current results for AN are different from the results reported for the whole category of EDs in the meta-analysis published in 2007 (Gratacòs et al. 2007). There, the ES was much higher and significant, with OR of 1.36 (95%CI; 1.18-1.57) (for the dominant model of the Met allele). This association signal might have been driven predominantly by EDs other than AN, i.e. bulimia nervosa, and ED not otherwise specified. A meta-analysis of those phenotypes – in separation from AN – is warranted. A promising (but practically challenging) approach to it is to use the classifications based on latent classes, latent profiles or taxometric analyses, rather than arbitrarily chosen sub- and intermediate phenotypes (Wonderlich et al. 2007; Eddy et al. 2009; Williamson et al. 2002). This strategy has already been employed with some success to the 5HTTLPR polymorphism in EDs (Steiger et al. 2009). The present results do not exclude the possibility that genetic variation at the BDNF locus contributes to development of AN. It is still possible that the association lies within a range of very small effect sizes (OR<1.1), and that it was not detectable with the present statistical power. A sample size of over 30k subjects would be necessary to achieve 80% power with OR of 1.071 (this ES was estimated in the second analysis). Furthermore, possibilities of more complex scenarios of association should be kept in mind. Epistatic and gene x environment interactions have not been addressed in the studies of AN and BDNF, and examples of such interactions
146 Chapter 6 in the different fields are abundant. For instance, 2 studies found no main effect of Val66Met on neuroticism scores (NEO-PI-R (Costa and McCrae 1995)), but people with the Met allele and at least one copy of the DAT 9- repeat allele had lower neuroticism and harm avoidance (Hunnerkopf et al. 2007), and carriers of the Met allele in combination with 5-HTTLPR LL allele scored higher on neuroticism (Terracciano et al. 2010) (in the same study the Met allele had an increasing effect on introversion). These examples illustrate the difficulty of finding a single SNP association, which may be obscured by the fact that the direction of the Met allele’s effect might be modulated by other genetic variants. Moreover, gene x environment interactions are suspected to contribute substantially to the variance of mental illnesses (Uher 2009; Campbell et al. 2010). A gene x environment interaction that is not accounted for may greatly reduce the power to detect the association. Finally, rather than with a single SNP, the disease may be associated with certain haplotypes within the BDNF locus. One of the weaknesses of meta-analyses is that it often has to sacrifice some phenotypic specificity of individual studies to be able to combine many of them. Due to the insufficient information in 3 studies (and hence decreased sample size), we did not distinguish between AN subtypes (restricting and bingeing/purging) in the main analysis. The results of an exploratory analysis on AN subtypes are available in the Supplementary Table S5. Another limitation is the fact that only 9 studies were included in the meta- analysis. This number did not allow for analyses of potential moderators, such as ethnicity or sex. Nevertheless, almost all cases included in the meta- analysis were female, thus sex, as a confounder, should not play a major role. Furthermore, only one of the included studies was performed on Asian participants (weight of the study was 3.7%) whereas the rest included predominantly Caucasian subjects. In conclusion, the present meta-analysis included 8 studies from literature and new genotype data from patients with AN and healthy controls from Utrecht (the Netherlands). The quality of analyzed evidence was good
147 Chapter 6 and the study was relatively well-powered. The results showed that the supposed association between rs6265 and AN became non-significant when the first published study was excluded (or replaced by a trio study on a partially overlapping case set). This association has been considered as one of the more robust findings in the genetic association studies of AN, but we could not confirm it in the present meta-analysis.
Acknowledgements
Marek K. Brandys was supported by funding from the Marie Curie Research Training Network INTACT (Individually tailored stepped care for women with eating disorders; reference number: MRTN-CT-2006-035988). The authors thank the Price Foundation and the Price Foundation Collaborative Group for data collection, genotyping, and data analysis. The authors are indebted to the participating families for their contribution of time and effort in support of this study.
References
American Psychiatric Association. 2000. Diagnostic and statistical manual of mental disorders, fourth edition, text revision. Washington, DC: American Psychiatric Association. Beckers S, Peeters A, Zegers D, Mertens I, Gaal LV, Van Hul W. 2008. Association of the BDNF Val66Met variation with obesity in women. Mol Genet Metab 95:110-112. Berton O, McClung CA, DiLeone RJ, Krishnan V, Renthal W, Russo SJ, et al. 2006. Essential role of BDNF in the mesolimbic dopamine pathway in social defeat stress. Science 311:864-868. Birmingham CL, Su J, Hlynsky JA, Goldner EM, Gao M. 2005. The mortality rate from anorexia nervosa. Int J Eat Disord 38:143-146.
148 Chapter 6
Brandys MK, Kas MJH, van Elburg AA, Campbell IC, Adan RAH. 2011. A meta-analysis of circulating BDNF concentrations in anorexia nervosa. World J Biol Psychiatry. In press. Brandys MK, van Elburg AA, Loos RJ, Bauer F, Hendriks J, van der Schouw YT, et al. 2009. Are recently identified genetic variants regulating BMI in the general population associated with anorexia nervosa? Am J Med Genet B Neuropsychiatr Genet 153B:695-699. Bulik CM, Thornton LM, Root TL, Pisetsky EM, Lichtenstein P, Pedersen NL. 2010. Understanding the relation between anorexia nervosa and bulimia nervosa in a swedish national twin sample. Biol Psychiatry 67:71- 77. Campbell IC, Mill J, Uher R, Schmidt U. 2010. Eating disorders, gene- environment interactions and epigenetics. Neurosci Biobehav Rev. Chen Z, Jing D, Bath KG, Ieraci A, Khan T, Siao C, et al. 2006. Genetic variant BDNF (Val66Met) polymorphism alters anxiety-related behavior. Science 314:140-143. Chiaruttini C, Vicario A, Li Z, Baj G, Braiuca P, Wu Y, et al. 2009. Dendritic trafficking of BDNF mRNA is mediated by translin and blocked by the G196A (Val66Met) mutation. Proceedings of the National Academy of Sciences 106:16481-16486. Cirulli F, Alleva E. 2009. The NGF saga: From animal models of psychosocial stress to stress-related psychopathology. Front Neuroendocrinol 30:379-395. Cooper Z, Fairburn C. 1987. The eating disorder examination: A semi- structured interview for the assessment of the specific psychopathology of eating disorders. Int J Eat Disord 6:1-8. Costa PT, McCrae RR. 1995. Domains and facets: Hierarchical personality assessment using the revised NEO personality inventory. J Pers Assess 64:21-50. Dardennes RM, Zizzari P, Tolle V, Foulon C, Kipman A, Romo L, et al. 2007. Family trios analysis of common polymorphisms in the obestatin/ghrelin, BDNF and AGRP genes in patients with anorexia nervosa:
149 Chapter 6
Association with subtype, body-mass index, severity and age of onset. Psychoneuroendocrinology, 32:106-113. de Krom M, Bakker SC, Hendriks J, van Elburg A, Hoogendoorn M, Verduijn W, et al. 2005. Polymorphisms in the brain-derived neurotrophic factor gene are not associated with either anorexia nervosa or schizophrenia in dutch patients. Psychiatr Genet 15:81-81. Eddy KT, Crosby RD, Keel PK, Wonderlich SA, le Grange D, Hill L, et al. 2009. Empirical identification and validation of eating disorder phenotypes in a multisite clinical sample. J Nerv Ment Dis 197:41-49. Egan MF, Kojima M, Callicott JH, Goldberg TE, Kolachana BS, Bertolino A, et al. 2003. The BDNF val66met polymorphism affects activity- dependent secretion of BDNF and human memory and hippocampal function. Cell 112:257-269. Friedel S, Horro FF, Wermter AK, Geller F, Dempfle A, Reichwald K, et al. 2005. Mutation screen of the brain derived neurotrophic factor gene (BDNF): Identification of several genetic variants and association studies in patients with obesity, eating disorders, and attention-deficit/hyperactivity disorder. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics 132B:96-99. Frustaci A, Pozzi G, Gianfagna F, Manzoli L, Boccia S. 2008. Meta- analysis of the brain-derived neurotrophic factor gene (BDNF) Val66Met polymorphism in anxiety disorders and anxiety-related personality traits. Neuropsychobiology 58:163-170. Gratacòs M, González JR, Mercader JM, de Cid R, Urretavizcaya M, Estivill X. 2007. Brain-derived neurotrophic factor Val66Met and psychiatric disorders: Meta-analysis of case-control studies confirm association to substance-related disorders, eating disorders, and schizophrenia. Biological Psychiatry 61:911-922. Gunstad J, Schofield P, Paul RH, Spitznagel MB, Cohen RA, Williams LM, et al. 2006. BDNF Val66Met polymorphism is associated with body mass index in healthy adults. Neuropsychobiology 53:153-156.
150 Chapter 6
Higgins JPT, Thompson SG, Deeks JJ, Altman DG. 2003. Measuring inconsistency in meta-analyses. BMJ 327:557-560. Hunnerkopf R, Strobel A, Gutknecht L, Brocke B, Lesch KP. 2007. Interaction between BDNF Val66Met and dopamine transporter gene variation influences anxiety-related traits. Neuropsychopharmacology 32:2552-2560. Hyman C, Hofer M, Barde Y, Juhasz M, Yancopoulos GD, Squinto SP, et al. 1991. BDNF is a neurotrophic factor for dopaminergic neurons of the substantia nigra. Nature 350:230-232. Kawashima K, Ikeda M, Kishi T, Kitajima T, Yamanouchi Y, Kinoshita Y, et al. 2009. BDNF is not associated with schizophrenia: Data from a japanese population study and meta-analysis. Schizophr Res 112:72-79. Kazeem GR, Farrall M. 2005. Integrating case-control and TDT studies. Ann Hum Genet 69:329-335. Koizumi H, Hashimoto K, Itoh K, Nakazato M, Shimizu E, Ohgake S, et al. 2004. Association between the brain-derived neurotrophic factor 196G/A polymorphism and eating disorders. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics 127B:125-127. Kortegaard LS, Hoerder K, Joergensen J, Gillberg C, Kyvik KO. 2001. A preliminary population-based twin study of self-reported eating disorder. Psychol Med 31:361-365. Lau J, Ioannidis JP, Terrin N, Schmid CH, Olkin I. 2006. The case of the misleading funnel plot. BMJ 333:597-600. Lommatzsch M, Zingler D, Schuhbaeck K, Schloetcke K, Zingler C, Schuff-Werner P, et al. 2005. The impact of age, weight and gender on BDNF levels in human platelets and plasma. Neurobiol Aging 26:115-123. Mantel N, Haenszel W. 1959. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 22:719-748. Montag C, Basten U, Stelzel C, Fiebach CJ, Reuter M. 2010. The BDNF Val66Met polymorphism and anxiety: Support for animal knock-in-studies from a genetic association study in humans. Psychiatry Res.
151 Chapter 6
Munafo MR, Flint J. 2004. Meta-analysis of genetic association studies. Trends Genet 20:439-444. Nakaoka H, Inoue I. 2009. Meta-analysis of genetic association studies: Methodologies, between-study heterogeneity and winner's curse. J Hum Genet 54:615-623. Narita M, Aoki K, Takagi M, Yajima Y, Suzuki T. 2003. Implication of brain-derived neurotrophic factor in the release of dopamine and dopamine- related behaviors induced by methamphetamine. Neuroscience 119:767- 775. Nicodemus K. 2008. Catmap: Case-control and TDT meta-analysis package. BMC Bioinformatics 9:130. Papadopoulos FC, Ekbom A, Brandt L, Ekselius L. 2009. Excess mortality, causes of death and prognostic factors in anorexia nervosa. The British Journal of Psychiatry 194:10-17. Pedersen BK, Pedersen M, Krabbe KS, Bruunsgaard H, Matthews VB, Febbraio MA. 2009. Role of exercise-induced BDNF production in the regulation of energy homeostasis. Experimental Physiology. Petryshen TL, Sabeti PC, Aldinger KA, Fry B, Fan JB, Schaffner SF, et al. 2009. Population genetic study of the brain-derived neurotrophic factor (BDNF) gene. Mol Psychiatry. Pezawas L, Verchinski BA, Mattay VS, Callicott JH, Kolachana BS, Straub RE, et al. 2004. The brain-derived neurotrophic factor val66met polymorphism and variation in human cortical morphology. J Neurosci 24:10099-10102. Pinheiro AP, Bulik CM, Thornton LM, Sullivan PF, Root TL, Bloss CS, et al. 2010. Association study of 182 candidate genes in anorexia nervosa. Am J Med Genet B Neuropsychiatr Genet. Purcell S, Cherny SS, Sham PC. 2003. Genetic power calculator: Design of linkage and association genetic mapping studies of complex traits. Bioinformatics 19:149-150. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MAR, Bender D, et al. 2007. PLINK: A tool set for whole-genome association and population-
152 Chapter 6 based linkage analyses. The American Journal of Human Genetics 81:559- 575. Ribases M, Gratacos M, Fernandez-Aranda F, Bellodi L, Boni C, Anderluh M, et al. 2004a. Association of BDNF with anorexia, bulimia and age of onset of weight loss in six european populations. Hum Mol Genet 13:1205- 1212. Ribases M, Gratacos M, Fernandez-Aranda F, Bellodi L, Boni C, Anderluh M, et al. 2004b. Association of BDNF with restricting anorexia nervosa and minimum body mass index: A family-based association study of eight european populations. Eur J Hum Genet 13:428-434. Rybakowski F, Dmitrzak-Weglarz M, Szczepankiewicz A, Skibinska M, Slopien A, Rajewski A, et al. 2007. Brain derived neurotrophic factor gene Val66Met and -270C/T polymorphisms and personality traits predisposing to anorexia nervosa. Neuro Endocrinol Lett 28:153-158. Sánchez-Mora C, Ribasés M, Ramos-Quiroga JA, Casas M, Bosch R, Boreatti-Hümmer A, et al. 2009. Meta-analysis of brain-derived neurotrophic factor p.Val66Met in adult ADHD in four european populations. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics 9999:. Schwarzer G. 2007. Meta: An R package for meta-analysis. R News 7:40-45. Shugart YY, Chen L, Day IN, Lewis SJ, Timpson NJ, Yuan W, et al. 2009. Two british women studies replicated the association between the Val66Met polymorphism in the brain-derived neurotrophic factor (BDNF) and BMI. Eur J Hum Genet 17:1050-1055. Slof-Op 't Landt MCT, Meulenbelt I, Bartels M, Suchiman E, Middeldorp CM, Houwing-Duistermaat JJ, et al. 2010. Association study in eating disorders: TPH2 associates with anorexia nervosa and self-induced vomiting. Genes Brain Behav. Stefansson H, Ophoff RA, Steinberg S, Andreassen OA, Cichon S, Rujescu D, et al. 2009. Common variants conferring risk of schizophrenia. Nature 460:744-747.
153 Chapter 6
Steiger H, Richardson J, Schmitz N, Joober R, Israel M, Bruce KR, et al. 2009. Association of trait-defined, eating-disorder sub-phenotypes with (biallelic and triallelic) 5HTTLPR variations. J Psychiatr Res. Strober M, Freeman R, Lampert C, Diamond J, Kaye W. 2000. Controlled family study of anorexia nervosa and bulimia nervosa: Evidence of shared liability and transmission of partial syndromes. Am J Psychiatry 157:393-401. Szeszko PR, Lipsky R, Mentschel C, Robinson D, Gunduz-Bruce H, Sevy S, et al. 2005. Brain-derived neurotrophic factor val66met polymorphism and volume of the hippocampal formation. Mol Psychiatry 10:631-636. Terracciano A, Tanaka T, Sutin AR, Deiana B, Balaci L, Sanna S, et al. 2010. BDNF Val66Met is associated with introversion and interacts with 5- HTTLPR to influence neuroticism. Neuropsychopharmacology 35:1083-1089. Thorleifsson G, Walters GB, Gudbjartsson DF, Steinthorsdottir V, Sulem P, Helgadottir A, et al. 2009. Genome-wide association yields new sequence variants at seven loci that associate with measures of obesity. Nat Genet 41:18-24. Uher R. 2009. The role of genetic variation in the causation of mental illness: An evolution-informed framework. Mol Psychiatry 14:1072-1082. Verhagen M, van der Meij A, van Deurzen PAM, Janzing JGE, Arias- Vasquez A, Buitelaar JK, et al. 2008. Meta-analysis of the BDNF Val66Met polymorphism in major depressive disorder: Effects of gender and ethnicity. Mol Psychiatry. Wade TD, Bulik CM, Neale M, Kendler KS. 2000. Anorexia nervosa and major depression: Shared genetic and environmental risk factors. Am J Psychiatry 157:469-471. Waterhouse EG, Xu B. 2009. New insights into the role of brain- derived neurotrophic factor in synaptic plasticity. Molecular and Cellular Neuroscience 42:.
154 Chapter 6
Williamson DA, Womble LG, Smeets MAM, Netemeyer RG, Thaw JM, Kutlesic V, et al. 2002. Latent structure of eating disorder symptoms: A factor analytic and taxometric investigation. Am J Psychiatry 159:412-418. Wonderlich SA, Joiner TE, Keel PK, Williamson DA, Crosby RD. 2007. Eating disorder diagnoses: Empirical approaches to classification. Am Psychol 62:167-180. Zintzaras E, Lau J. 2008. Synthesis of genetic association studies for pertinent gene–disease associations requires appropriate methodological and statistical approaches. J Clin Epidemiol 61:634-645.
Supplementary data
General
Supplementary Figure S1. Flow diagram.
Supplementary Figure S1. Study selection diagram. Search terms: (bdnf OR brain derived OR val66met OR val/met OR rs6265 OR 196G/A) AND (anorexia OR eating disorders) AND (association OR gene-association OR polymorphism).
155 Chapter 6
First analysis: the allelic contrast and the dominant effect
Supplementary Table S1. Influence table, 9 studies.
Supplementary Table S1. Influence analysis for 8 case-control studies and 1 family- based; testing the allelic contrast. Each row shows summary results when an indicated study is omitted. The pooled ORs in a fixed-effect model. Omitted study Studies OR 95% Conf. Int. P I2 Omitting Ribases 2004a k=8 1.06 [0.96-1.18] 0.25 0% Omitting Koizumi 2004 k=8 1.12 [1.01-1.23] 0.02 0% Omitting Friedel 2005 k=8 1.12 [1.02-1.23] 0.02 0% Omitting de Krom 2005 k=8 1.11 [1.01-1.23] 0.04 0% Omitting Dardennes 2007 k=8 1.11 [1.01-1.22] 0.03 0% Omitting Rybakowski 2007 k=8 1.11 [1.01-1.22] 0.03 0% Omitting Pinheiro 2010 k=8 1.14 [1.02-1.28] 0.02 0% Omitting Slof-Op ’t Landt 2011 k=8 1.13 [1.02-1.24] 0.02 0% Omitting Brandys 2011 k=8 1.12 [1.02-1.24] 0.02 0% Pooled estimate k=9 1.11 [1.01-1.22] 0.02 0%
Supplementary Figure S2. Forest plot, dominant effect, 8 studies.
Supplementary Figure S2. Forest plot presenting ORs for anorexia nervosa, for the dominant effect of the Met allele. 8 case-control studies are included. The weight of each study is reflected by the size of squares, and whiskers represent 95% confidence intervals. The pooled OR is based on the fixed effect model. I2, as a measure of heterogeneity, equals 0%.
156 Chapter 6
Supplementary Figure S3. Reversed cumulative plot, dominant effect, 8 studies.
Supplementary Figure S3. Cumulative meta-analysis for the dominant effect of the Met allele. 8 case-control studies are included. Studies are added in a reversed chronological order; each row represents the pooled OR for all studies added to this point (based on the fixed-effect model). The whiskers represent (cumulative) 95% confidence intervals.
Supplementary Table S2. Influence table, dominant effect, 8 studies.
Supplementary Table S2. Influence analysis for 8 case-control studies; testing a dominant effect of the Met allele. Each row shows summary results when an indicated study is omitted. The pooled ORs in a fixed-effect model. I2 as a measure of heterogeneity. Omitted study Studies OR 95% Conf. Int. P I2 Omitting Ribases 2004a k=7 1.07 [0.94-1.21] 0.32 0% Omitting Koizumi 2004 k=7 1.13 [1-1.26] 0.04 7.1% Omitting Friedel 2005 k=7 1.15 [1.03-1.29] 0.02 0.1% Omitting de Krom 2005 k=7 1.15 [1.02-1.3] 0.02 13% Omitting Rybakowski 2007 k=7 1.13 [1.01-1.27] 0.04 10.4% Omitting Pinheiro 2010 k=7 1.17 [1.02-1.35] 0.02 8.7% Omitting Slof-Op ’t Landt 2011 k=7 1.16 [1.03-1.3] 0.02 5.9% Omitting Brandys 2011 k=7 1.15 [1.02-1.3] 0.03 14.9% Pooled estimate k=8 1.14 [1.02-1.27] 0.02 2.8%
157 Chapter 6
Second analysis: the allelic contrast and the dominant effect
Supplementary Table S3. Influence table, with the trio study.
Supplementary Table S3. Influence analysis for 7 case-control studies and 2 family- based (the case-control study by Ribases et al. 2004a replaced with the trio study by Ribases et al. 2004b); testing the allele contrast. Each row shows summary results when an indicated study is omitted. The pooled ORs in a fixed-effect model. Omitted study Studies OR 95% Conf. Int. P I2 Omitting Koizumi 2004 k=8 1.07 [0.97-1.18] 0.17 0% Omitting Ribases 2004b k=8 1.06 [0.96-1.18] 0.25 0% Omitting Friedel 2005 k=8 1.08 [0.98-1.19] 0.14 0% Omitting de Krom 2005 k=8 1.06 [0.96-1.18] 0.23 0% Omitting Dardennes 2007 k=8 1.07 [0.97-1.18] 0.17 0% Omitting Rybakowski 2007 k=8 1.07 [0.97-1.18] 0.17 0% Omitting Pinheiro 2010 k=8 1.07 [0.97-1.18] 0.2 0% Omitting Slof-Op ’t Landt 2011 k=8 1.08 [0.96-1.22] 0.18 0% Omitting Brandys 2011 k=8 1.08 [0.98-1.19] 0.14 0% Pooled estimate k=9 1.07 [0.97-1.18] 0.16 0%
Supplementary Figure S4. Forest plot, 7 studies.
Supplementary Figure S4. Forest plot presenting ORs for the dominant effect of the Met allele, without the studies by Ribases et al. 2004a,b (either case-control or the family-based). 7 Case-control studies only. The weight of each study is reflected by the size of squares, and whiskers represent 95% confidence intervals. The pooled OR is based on the fixed effect model. I2, as a measure of heterogeneity, equals 0%.
Supplementary Figure S5. Reversed cumulative plot, dominant effect, 7 studies.
158 Chapter 6
Supplementary Figure S5. Cumulative meta-analysis for the dominant effect of the Met allele, without the studies by Ribases et al. 2004a,b (either case-control or the family-based). 7 Case-control studies only. Studies are added in a reversed chronological order; each row represents the pooled OR for all studies added to this point (based on the fixed-effect model and for the allelic contrast). ). The whiskers represent (cumulative) 95% confidence intervals. I2 as a measure of heterogeneity was 0% at every step.
Supplementary Table S4. Influence table, dominant effect, 7 studies.
Supplementary Table S4. Influence analysis for 7 case-control studies; testing the dominant effect of the Met allele. Each row shows summary results when an indicated study is omitted. The pooled ORs in a fixed-effect model. Omitted study Studies OR 95% Conf. Int. P I2 Omitting Koizumi 2004 k=6 1.05 [0.92-1.2] 0.46 0% Omitting Friedel 2005 k=6 1.08 [0.95-1.23] 0.24 0% Omitting de Krom 2005 k=6 1.07 [0.93-1.23] 0.33 0% Omitting Rybakowski 2007 k=6 1.05 [0.92-1.2] 0.44 0% Omitting Pinheiro 2010 k=6 1.07 [0.9-1.27] 0.43 0% Omitting Slof-Op ’t Landt 2011 k=6 1.08 [0.94-1.24] 0.26 0% Omitting Brandys 2011 k=6 1.07 [0.93-1.23] 0.38 0% Pooled estimate k=7 1.07 [0.94-1.21] 0.32 0%
Exploratory analysis
Supplementary Table S5. Exploratory analysis of anorexia nervosa subtypes
Supplementary Table S5. Results of the exploratory analysis on anorexia nervosa subtypes (k=6). Data on subtypes were not available in 3 studies. Controls were the same as in the main analyses. Providing the heterogeneity was larger than zero, the
159 Chapter 6 analyzes were performed under two models meta-analysis (fixed effect and random effects). Each subtype was analyzed under two scenarios: with the case-control study from Ribases et al. 2004a or with the family-based study by Ribasese et al. 2004b. N Group; Meta- N N I- Pooled 95% Conf. Inf. P analysis model cases controls square OR Int. Trios ANR; 1014 2560 30 8.85% 1.16 [1.02-1.32] 0.03 Fixed effecta ANR; - - - - 1.16 [1.01-1.33] 0.04 Random effectsa ANR (Trio); 553 1407 188 0.00% 1.12 [0.97-1.28] 0.12 Fixed effectb ANBP; 1174 2560 23 44.62% 1.09 [0.96-1.24] 0.21 Fixed effecta ANBP; - - - - 1.04 [0.85-1.27] 0.7 Random effectsa ANBP (Trio); 749 1407 136 10.18% 0.99 [0.87-1.14] 0.93 Fixed effectb ANBP (Trio); - - - - 0.98 [0.84-1.15] 0.79 Random effectsb a included studies: Ribases et al. 2004a, Koizumi et al. 2004, Dardennes et al. 2007, Pinheiro et al. 2010, Slof-Op 't Landt et al. 2011, Brandys et al. 2011; b included studies: Koizumi et al. 2004, Ribases et al. 2004b, Dardennes et al. 2007, Pinheiro et al. 2010, Slof-Op 't Landt et al. 2011, Brandys et al. 2011; ANR – anorexia nervosa restricting; ANBP – anorexia nervosa bingeing/purging. According to the results of these exploratory analyses, the ANR subtype appears to be marginally associated with rs6265. However, the association signal is predominantly driven by the case-control study from Ribases et al. 2004a. The signal dissipates if this study is replaced by the family-based one performed on an overlapping sample (from the same group, Ribases et al. 2004b). This follows exactly the same pattern as in the main analysis, with decreased power, however. The ANBP subtype displays no indication of association and the heterogeneity of effect sizes in this group is higher than in the ANR.
160 Chapter 6
PRISMA Checklist
Reported Section/topic # Checklist item on page #
TITLE Title 1 Identify the report as a systematic review, meta-analysis, or both. 1
ABSTRACT Structured summary 2 Provide a structured summary including, as applicable: background; objectives; data sources; study eligibility criteria, 3 participants, and interventions; study appraisal and synthesis methods; results; limitations; conclusions and implications of key findings; systematic review registration number.
INTRODUCTION
Rationale 3 Describe the rationale for the review in the context of what is already known. 4-7
Objectives 4 Provide an explicit statement of questions being addressed with reference to participants, interventions, comparisons, 7 outcomes, and study design (PICOS).
METHODS
Protocol and registration 5 Indicate if a review protocol exists, if and where it can be accessed (e.g., Web address), and, if available, provide N/A registration information including registration number.
Eligibility criteria 6 Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered, 8 language, publication status) used as criteria for eligibility, giving rationale.
Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional 8-9 studies) in the search and date last searched.
Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated. Suppl. Fig. 1
Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, 8, Suppl. included in the meta-analysis). Fig. 1
161 Chapter 6
Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for 8-9 obtaining and confirming data from investigators.
Data items 11 List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and 8-9 simplifications made.
Risk of bias in individual 12 Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at 9-10,12 studies the study or outcome level), and how this information is to be used in any data synthesis.
Summary measures 13 State the principal summary measures (e.g., risk ratio, difference in means). 9
Synthesis of results 14 Describe the methods of handling data and combining results of studies, if done, including measures of consistency (e.g., 9-10 I2) for each meta-analysis.
Reported Section/topic # Checklist item on page #
Risk of bias across 15 Specify any assessment of risk of bias that may affect the cumulative evidence (e.g., publication bias, selective reporting 9-10,12 studies within studies).
Additional analyses 16 Describe methods of additional analyses (e.g., sensitivity or subgroup analyses, meta-regression), if done, indicating which 9-10,12 were pre-specified.
RESULTS
Study selection 17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each 11, Suppl. stage, ideally with a flow diagram. Fig. 1
Study characteristics 18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and Table 1 & 2 provide the citations.
Risk of bias within 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12). 12 studies
162 Chapter 6
Results of individual 20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention 11, Fig. 3, studies group (b) effect estimates and confidence intervals, ideally with a forest plot. Suppl. Table 1
Synthesis of results 21 Present results of each meta-analysis done, including confidence intervals and measures of consistency. 11-15, Fig. 2 & 4
Risk of bias across 22 Present results of any assessment of risk of bias across studies (see Item 15). 12-15, Fig. 3 studies & 5
Additional analysis 23 Give results of additional analyses, if done (e.g., sensitivity or subgroup analyses, meta-regression [see Item 16]). Fig. 3 & 5, Suppl. Table 1-4, Supp. Fig. 2-5 DISCUSSION Summary of evidence 24 Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key 15-16 groups (e.g., healthcare providers, users, and policy makers).
Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified 17-18 research, reporting bias).
Conclusions 26 Provide a general interpretation of the results in the context of other evidence, and implications for future research. 16-18
FUNDING Funding 27 Describe sources of funding for the systematic review and other support (e.g., supply of data); role of funders for the 24 systematic review.
From: Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group (2009). Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLoS Med 6(6): e1000097. doi:10.1371/journal.pmed1000097 For more information, visit: www.prisma-statement.org.
163 Chapter 7
Chapter 7
No evidence for involvement of CNVs associated with neurodevelopmental disorders in anorexia nervosa
Marek K. Brandys Bobby Koeleman Carolien de Kovel Genetic Consortium for Anorexia Nervosa Cynthia M. Bulik David A. Collier Roger A. H. Adan
Unpublished manuscript
The data of the control groups, which were available to us at the time of performing this study, came from various Illumina genotyping platforms and were at different stages of processing (Illumina intensity data, raw CNV calls, QC-ed CNV calls). For these reasons, the results have been deemed preliminary (see the limitations section of the manuscript) and were not submitted for publication. In the meantime and recently, a study which used the same data of patients with AN and had access to other control data (dbGAP controls) came up with similar results, i.e. lack of association between selected rare CNVs and AN (Yilmaz et al., 2015, abstract at the World Congress of Psychiatric Genetic).
164 Chapter 7
Abstract
Several large and rare copy number variants (CNVs) have been shown to be associated with neurodevelopmental disorders, such as mental disability, epilepsy, schizophrenia, and autism, as well as obesity (whether obesity is neurodevelopmental can be a matter of debate). It is unknown if these pleiotropic CNVs play a role in anorexia nervosa (AN). AN is a putatively neurodevelopmental psychiatric disorder, characterised by extremely low body-weight. It is known for high mortality rate and a substantial social burden. We performed an association analysis of twelve large and rare CNVs, which have been shown to be associated with neurodevelopmental and psychiatric disorders and obesity, in a dataset comprising 2959 cases with AN and 9101 controls. Genotyping of cases was performed as a part of a genome-wide association study of AN1. This sample has 80% power to detect association of the selected CNV if it confers a relative risk of 3.8 for AN, an effect smaller than that observed for the selected CNVs in other psychiatric disorders. CNV calls were made by PennCNV in all cases and controls and were visually inspected and verified by QuantiSNP in the case group and in 3 out of 6 control groups. Forty-seven CNVs were detected in the selected candidate regions in cases (frequency of 1.81%) and 236 in controls (frequency of 2.59%). However, no single CNV was significantly more or less frequent in cases than controls, after correction for multiple testing. The findings in this largest study of candidate CNVs in AN do not yield support for the hypothesis that candidate CNVs, which were previously associated with neurodevelopmental disorders, associate with AN. In view of several important limitations, these results should be considered tentative.
165 Chapter 7
Introduction
Anorexia nervosa (AN) is chronic and serious eating disorder: known for the highest standardized mortality rate of all psychiatric illnesses (mortality rate is 6-10 times higher than in a reference population 2,3), AN is difficult to treat, and its etiology remains unclear. Between 46 and 78% of variance in liability to AN is attributable to additive genetic factors 4-6, and family studies have determined a ten-fold increase in lifetime risk of developing AN for a first- degree female relative of a proband with AN (compared to relatives of unaffected individuals) 7. Similarities between AN and several disorders such as autism spectrum disorders (ASD) 8, schizophrenia (SCZ) 9,10, obsessive-compulsive disorder 11 or obesity (where AN and obesity lie on the opposite ends of the weight spectrum12,13) have been reported, and it has been implied that the etiology of these conditions may share common pathways 14,15. Interestingly, evidence suggesting shared neurodevelopmental pathogenesis at the level of copy number variation (CNV) has been found not only for neuropsychiatric disorders, but also for seemingly unrelated phenotypes such as idiopathic generalized epilepsy (IGE) 16. Several large, rare, and recurrent (i.e., occurring in the same region of the genome in different individuals) CNVs have been found to be associated with intellectual disability, developmental delay, schizophrenia, bipolar disorder, IGE, and obesity 17-19. Whether these pleiotropic CNVs play a role in AN remains to be established. The study of the genetics of AN is nascent. Most candidate gene studies reported on small sample sizes, failed to replicate, or yielded null findings 20. One published genome-wide association study (GWAS) of AN 21 used the Illumina HumanHap610 genotyping platform. Underpowered as for the current standards, the study looked into single nucleotide polymorphism (SNP) frequencies of 1033 cases with AN and 3773 controls and identified no genome-wide significant hits. The authors also analysed selected CNVs, but given the relatively small sample size, their conclusions were limited.
166 Chapter 7
A larger GWAS of AN has been performed more recently 1, and it provides a basis for the present report. This GWA study created opportunity to investigate candidate CNVs in the largest sample of AN cases assembled to date. We explored the frequencies of candidate CNVs reported to be associated with central nervous system disorders and obesity, in cases with AN and several control groups. Additionally, we analysed two chromosomal regions which were highlighted in the previous AN GWAS study 21 as worthy of further investigation. Explicating the role of CNVs in AN has the potential to shed light on the pathogenesis of the disorder and its possible relation to other neurodevelopmental syndromes and obesity.
Subjects and methods
Subjects
The case group comprised females with a history of DSM-IV AN or eating disorder not otherwise specified AN-subtype (lifetime diagnosis; restricting or binge/purge subtype). The amenorrhea criterion was not required as it does not improve diagnostic specificity 22. Due to a high rate of diagnostic cross-overs, a lifetime occurrence of bulimia nervosa was not an excluding factor 23. The exclusion criteria included a diagnosis of conditions possibly confounding the diagnosis of AN (such as mental retardation or psychotic disorders). The samples originate from Europe and USA, but they were all of European descent (see Supplementary Material for details). Participants signed an informed consent to take part in genetic research, in accordance with the Declaration of Helsinki. The genotyping was performed in The Wellcome Trust Sanger Institute in London, United Kingdom on the Illumina 660W-Quad arrays (Illumina, Inc., San Diego, CA, USA). A total of 2959 cases which passed the GWAS quality control (QC) were included in the CNV-detection step of the analysis (see Supplementary Material for details). QC was done according to guidelines in 24.
167 Chapter 7
In addition to the AN cases, six control groups were included in this study (Table 1). Controls were not screened for AN; however, given low prevalence of AN in the general population (0.3% prevalence in young females 59), this represents a small and conservative bias. The controls were not age- or sex-matched, and they were all of European descent. The Greek (n=79) and the Finnish (n=399) controls were genotyped together with the cases, whereas additional control groups were culled from previously genotyped samples (Table 1). Illumina Final Report files were available for cases and Greek, Finnish, and Dutch controls. A file with raw CNV calls was available for the Wellcome Trust Case Control Consortium, Second Wave (WTCCC2) controls (n=5177) 25. QC procedures were applied to the above groups uniformly (as described below). CNV calls in the TwinsUK (n=1641) dataset came QC-ed by the provider 26. The data of the Children's Hospital of Philadelphia (CHOP) controls (European subjects only; n=1320) were extracted from a publicly available online CNV viewer (http://cnv.chop.edu), and were also QC-ed by the provider (this group overlaps partially with a control group used in 21).
168 Chapter 7
Table 1. Details of case and control groups CNV Group N Sex Ethnicity Genotyping Notes detection Reference algorithm European PennCNV Patients with femal Illumina 2829 descent and Original data AN (cases) es 660W-Quad only QuantiSNP genotyped PennCNV Greek femal Illumina 79 Greek along with and Original data (controls) es 660W-Quad the cases QuantiSNP genotyped PennCNV Finnish femal Illumina 399 Finnish along with and Original data (controls) es 660W-Quad the cases QuantiSNP PennCNV Dutch Illumina 579 mixed Dutch and 27 (controls) 550Hap QuantiSNP consensus 1958 British set of probes birth cohort WTCCC2 between and the 5177 mixed British PennCNV 25 (controls) Illumina610- national Q and blood service Illumina 1 cohort
TwinsUK - St. Unrelated Thomas Twin Illumina individuals http://twinsuk 1641 mixed British PennCNV Registry Human610-Q only were .ac.uk/; 26 (controls) used
customize d analysis the Children's workflow Hospital of European Illumina Healthy (Circulary http://cnv.cho Philadelphia 1320 mixed descent 550Hap children Binary p.edu/; 28 (CHOP; only Segmentat controls) ion algorithm) WTCCC2 - Wellcome Trust Case Control Consortium, Second Wave
CNV selection
CNV candidates were chosen from the literature, on the basis of their association with at least one neuropsychiatric disorder, namely intellectual disability, schizophrenia, autism, or epilepsy. We also selected CNV candidates that have been previously linked to obesity. Selected CNVs also needed to have dense probe coverage on the Illumina platforms used for genotyping of cases and controls. Coordinates, known associations and
169 Chapter 7 references per each CNV can be found in Table 2. These large CNVs with established coordinates are reliably called from high-density genotyping platforms 29. In addition, we checked two chromosomal regions highlighted by 21 as potentially interesting for further study in the context of AN.
Table 2. CNVs selected to be tested for association with anorexia nervosa
Chromosomal States Known Coordinates hg18 References band tested associations
chr1:144,000,001- SCZ, TAR 1q21.1 proximal dup 144,500,000 Syndrome 30,31 chr1:144,800,001- dup and 1q21.1 distal SCZ, DD, BP 146,350,000 del 29,32-35 chr2:50,000,001- 2p16.3 del SCZ, ASD 51,300,000 32,36,37 chr15:20,200,001- dup and 15q11.2 ASD, IGE, DD 20,800,000 del 16,29,31,33,34,36 chr15:28,300,000- dup and SCZ, ASD, ADHD, 15q13.3 30,700,000 del BP, ID, DD 29,32,34-36,38 chr16:14,900,001- dup and 16p13.11 ADHD, IGE, ID 16,400,000 del 16,39 chr16:28,650,001- dup and 16p11.2 distal ASD, SCZ, Ob 29,000,000 del 40,41 16p11.2 chr16:29,500,000- dup and ASD, ID, Ob proximal 30,150,000 del 18,29,32,33,36,42 chr17:13,900,000- dup and 17p12 ASD, SCZ 15,500,000 del 29,33,43 chr22:17,200,001- dup and 22q11.21 ASD, SCZ, ID 19,900,000 del 29,32,33,36 ASD, Smith- chr17:17,200,000- 17p11.2 del Magenis 18,000,000 syndrome, DD, Ob 19,44,45 chr1:9,780,0000- ASD, Speech 1p21.3 del 98,400,000 delay, ID, Ob 19,46,47 chr2:1,195,000- 2p25.3 del SCZ, ID, Ob 3,100,000 19,48,49 chr11:27,200,001- ADHD, DD, ASD, 11p14.1 del 31,000,000 Ob 19,50 chr3:800,000- dup and 3p26.3a 21 1,600,000 del
170 Chapter 7
chr13:22426685- dup and 13q12.12a 21 23795901 del a- from Wang et al. 2011 study on CNVs in anorexia nervosa; SCZ- schizophrenia; BP- bipolar disorder; ASD- autism spectrum disorder; IGE- idiopathic generalized epilepsy; DD- developmental delay; ADHD- attention deficit hyperactivity disorder; ID- intellectual disability; Ob- obesity
CNV detection
A hidden Markov model with the genomic wave adjustment (GC model; 51) algorithm implemented in the PennCNV package (2011Jun16 version; 52) was applied to Illumina Final Report files. This algorithm utilizes multiple sources of information, i.e. total intensity signal from each probe at a locus (LogRR, LogR ratio), relative intensity signal of probes at a locus (BAF, beta-allele frequency), the allele frequency of SNPs and the distance between adjacent SNPs. Since none of the candidate CNVs was located on the sex chromosomes, only the autosomes were analysed.
Quality control
The genome-wide CNV data used in the present study were subjected to several quality control and filtering procedures. Individuals with the standard deviation of the LogRR larger than 0.35, BAF drift larger than 0.002, waviness factor larger or smaller than 0.05 or total number of CNVs larger than 300 were removed (Supplementary Figures 1-4 show histograms with distribution of samples per each parameter; detailed description of these parameters can be found in 52)). Subsequent steps involved CNV-level QC and filtering. First, CNVs with a PennCNV confidence score lower than 10 were removed (this score represents the difference of the log likelihood of the most likely copy number state and the less likely copy number state). Since the calling algorithms may sometimes split large CNVs in two (especially if there is varying probe density in the region), a merging procedure was applied (if the
171 Chapter 7 gap between CNVs of the same state was less than 20% of the total length of these CNVs plus the gap, the calls were merged into one). Consequently, CNVs consisting of less than 10 probes were removed. Since the interest of the current study was only in the large, candidate CNVs, all the CNVs spanning less than 200kb were removed. Finally, all the CNVs lying outside the specified candidate regions were removed. Filtering and QC steps are shown in Table 3. TwinsUK data came in QC-ed (LogRR sd<0.35; SNP number per CNV > 10; merged with --fraction argument 50%; samples with CNV calls > 40 removed; since the data were from twin pairs, random twin from each pair was removed). In case of the CHOP sample, CNVs in the candidate regions, which were larger than 200kb, were extracted from the online database. These data were QC-ed by the provider (subjects with less than 98% genotyping rate and LogRR sd larger than 0.35 were excluded)
Table 3. Summary of QC and filtering steps and the resulting numbers of subjects or CNVs per each step
AN Greek Finnish Dutch WTCCC2 TwinsUK QC step CHOP cases controls controls controls controls controls sample Genotyping 3425 80 413 643 N/A N/A N/A size sample GWAS QC 2959 80 404 580 5621 N/A N/A size CNV detection CNV count 480407 10289 53932 13547 120042 N/A N/A sample Sample-level QC 2603 79 396 576 5089 1641 1320 size CNV count 401007 10289 51255 13440 69466 N/A N/A CNV QC CNV count 317137 8229 43155 4044 34240 N/A N/A CNV merging CNV count 316761 8216 43104 3997 33355 N/A N/A Selection of CNVs > CNV count 1790 56 191 678 4143 N/A N/A 200kb Number of CNVs > 200kb in the CNV count 65 4 4 11 153 31 46 candidate regions Number of CNVs > 200kb after CNV count 47 0 3 11 145 31 46 exclusion of 17p11.2 and 1p21.3 Number of CNVs confirmed by CNV count 47 0 3 11 N/A N/A N/A QuantiSNP in these regions
172 Chapter 7
Number of CNVs confirmed by visual CNV count 47 0 3 11 N/A N/A N/A inspection of LogRR and BAF graphs 17p11.2 and 1p21.3 regions were excluded due to lack of confirmation of PennCNV calls by QuantiSNP. Concordance rate was 100% in the remaining regions. Two regions from Wang et al. 2011 are not included. QC- quality control; AN- anorexia nervosa.
Where the intensity data were available, graphs presenting LogRR and BAF per each of the CNVs within the candidate regions were generated and visually inspected to exclude technical artefacts, as a result of which no CNV was excluded. The lowest number of probes contained in a CNV was 87. Altogether, 33 duplications, 34 hemizygous, and 0 homozygous deletions were found in cases. Finally, the CNV calls in cases and the control groups for which the intensity data were available were verified by the QuantiSNP algorithm. QuantiSNP uses an Objective Bayes Hidden-Markov Model to detect CNVs 53. All calls in the candidate regions made by PennCNV were confirmed by QuantiSNP, except for those on chr17p11.2 and 1p21.3. Owing to that, these two regions are not included in the further analyses (see Supplementary Material for additional info). The intensity data were not available for three of the six control groups, thus verification by QuantiSNP could not be applied. In the case group and the other three control groups, the PennCNV calls in the candidate regions were confirmed by QuantiSNP with 100% concordance (except for the two non-classical regions which were excluded from the analyses). The fact that the intensity data were not available for the large part of the control sample entails several limitations which are discussed in the final sections.
Statistical analysis
173 Chapter 7
Analyses of associations were carried out by means of two-sided Fisher’s exact tests. The resulting p-values were adjusted for multiple testing via Bonferroni correction. These calculations were made in Microsoft Excel 2007 and the power was calculated with the Genetic Power Calculator 54.
Results
Association analysis
Large CNVs detected at 12 pre-selected genomic regions were counted and tested for association with AN (there were 20 tests in total, since in some regions both deletions and duplications were considered). Table 4 presents the counts, frequencies, odds ratios with confidence intervals (where applicable) and p-values from two-sided Fisher’s tests for association. None of the CNVs associated with SCZ, BP or ASD (among other phenotypes) was significantly associated with AN, and the frequencies in controls reported in this study correspond with the frequencies described in the literature. Additionally, the counts of all CNVs, duplications, and deletions in the case and control groups are reported in Table 4. This comparison suggests slight underrepresentation of duplications in cases versus controls. Finally, we tested CNVs in two regions indicated in Wang et al. 2011. These regions were not associated with AN in the present study (Table 4).
174 Chapter 7
Table 4. Counts and results of the association analysis of the candidate CNVs in anorexia nervosa (CNVs > 200kb).
Cases Ctrl Ctrl Ctrl Ctrl Ctrl Ctrl N => 2603 79 396 576 5089 1641 1320 2603 9101 p-val Fisher's CNV -95% +95% corr. AN Greek Finnish Dutch WTCCC2 TwinsUK CHOP Sum Cases Sum Ctrl OR exact p- region CI CI for 20 value tests 1q21.1 Dup 4 (0,15%) 0 (0%) 0 (0%) 0 (0%) 5 (0,1%) 1 (0,06%) 0 (0%) 4 (0,15%) 6 (0,07%) 2,33 0,61 7,68 0,24 1 proximal 1q21.1 Dup 1 (0,04%) 0 (0%) 0 (0%) 0 (0%) 4 (0,08%) 1 (0,06%) 1 (0,08%) 1 (0,04%) 6 (0,07%) 0,58 0,07 4,5 1 1 distal Del 0 (0%) 0 (0%) 0 (0%) 0 (0%) 3 (0,06%) 0 (0%) 0 (0%) 0 (0%) 3 (0,03%) 0 0,11 10,42 1 1
2p16.3a Del 1 (0,04%) 0 (0%) 0 (0%) 0 (0%) 2 (0,04%) 1 (0,06%) 1 (0,08%) 1 (0,04%) 4 (0,04%) 0,87 0,09 7,27 1 1 1 4 15q11.2 Dup 7 (0,27%) 0 (0%) 18 (0,35%) 2 (0,12%) 3 (0,23%) 7 (0,27%) 28 (0,31%) 0,87 0,35 1,86 0,84 1 (0,25%) (0,69%) 2 Del 9 (0,35%) 0 (0%) 0 (0%) 20 (0,39%) 3 (0,18%) 5 (0,38%) 9 (0,35%) 30 (0,33%) 1,05 0,46 2,06 0,85 1 (0,35%) 1 15q13.3 Dup 14 (0,54%) 0 (0%) 0 (0%) 48 (0,94%) 13 (0,79%) 18 (1,36%) 14 (0,54%) 80 (0,88%) 0,61 0,32 1,01 0,1 1 (0,17%) Del 2 (0,08%) 0 (0%) 0 (0%) 0 (0%) 1 (0,02%) 1 (0,06%) 0 (0%) 2 (0,08%) 2 (0,02%) 3,49 0,81 29,17 0,22 1 1 3 16p13.11 Dup 3 (0,12%) 0 (0%) 9 (0,18%) 4 (0,24%) 2 (0,15%) 3 (0,12%) 19 (0,21%) 0,55 0,15 1,74 0,45 1 (0,25%) (0,52%) Del 1 (0,04%) 0 (0%) 0 (0%) 0 (0%) 4 (0,08%) 0 (0%) 0 (0%) 1 (0,04%) 4 (0,04%) 0,87 0,09 7,27 1 1 16p11.2 Dup 0 (0%) 0 (0%) 0 (0%) 0 (0%) 3 (0,06%) 0 (0%) 1 (0,08%) 0 (0%) 4 (0,04%) NA 0 NA 0,58 1 distal Del 0 (0%) 0 (0%) 0 (0%) 0 (0%) 2 (0,04%) 0 (0%) 1 (0,08%) 0 (0%) 3 (0,03%) NA 0 NA 1 1 16p11.2 Dup 0 (0%) 0 (0%) 0 (0%) 0 (0%) 2 (0,04%) 0 (0%) 1 (0,08%) 0 (0%) 3 (0,03%) NA 0 NA 1 1 proximal
175 Chapter 7
Del 1 (0,04%) 0 (0%) 0 (0%) 0 (0%) 5 (0,1%) 1 (0,06%) 1 (0,08%) 1 (0,04%) 7 (0,08%) 0,5 0,06 3,78 0,69 1
17p12 Dup 0 (0%) 0 (0%) 0 (0%) 0 (0%) 1 (0,02%) 1 (0,06%) 5 (0,38%) 0 (0%) 7 (0,08%) NA 0 NA 0,36 1 1 Del 0 (0%) 0 (0%) 0 (0%) 2 (0,04%) 1 (0,06%) 0 (0%) 0 (0%) 4 (0,04%) NA 0 NA 0,58 1 (0,25%) 1 22q11.21 Dup 4 (0,15%) 0 (0%) 0 (0%) 16 (0,31%) 2 (0,12%) 4 (0,3%) 4 (0,15%) 23 (0,25%) 0,61 0,2 1,64 0,49 1 (0,17%) Del 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 1 (0,08%) 0 (0%) 1 (0,01%) NA 0 NA 1 1
2p25.3 Del 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 1 (0,08%) 0 (0%) 1 (0,01%) NA 0 NA 1 1
11p14.1 Del 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 1 (0,08%) 0 (0%) 1 (0,01%) NA 0 NA 1 1 Total 3 11 145 236 47 (1,81%) 0 (0%) 31 (1,89%) 46 (3,48%) 47 (1,81%) 0,7 0,62 1,09 0,02 CNVs (0,76%) (1,91%) (2,85%) (2,59%) Total 2 9 106 176 33 (1,27%) 0 (0%) 24 (1,46%) 35 (2,65%) 33 (1,27%) 0,66 0,42 0,89 0,03 Dups (0,51%) (1,56%) (2,08%) (1,93%) 1 2 Total Dels 14 (0,54%) 0 (0%) 39 (0,77%) 7 (0,43%) 11 (0,83%) 14 (0,54%) 60 (0,66%) 0,82 0,88 2,18 0,58 (0,25%) (0,35%) 1 3p26.3b Dup 4 (0,15%) 0 (0%) 0 (0%) 4 (0,08%) 4 (0,24%) 2 (0,15%) 4 (0,15%) 11 (0,12%) 1,27 0,38 3,71 0,76 (0,17%) Del 2 (0,08%) 0 (0%) 0 (0%) 0 (0%) 3 (0,06%) 2 (0,12%) 0 (0%) 2 (0,08%) 5 (0,05%) 1,4 0,25 6,7 0,66 1 13q12.12b Dup 1 (0,04%) 0 (0%) 0 (0%) 9 (0,18%) 0 (0%) 0 (0%) 1 (0,04%) 10 (0,11%) 0,35 0,04 2,54 0,47 (0,17%) 1 Del 1 (0,04%) 0 (0%) 0 (0%) 1 (0,02%) 0 (0%) 0 (0%) 1 (0,04%) 2 (0,02%) 1,75 0,15 17,93 0,53 (0,17%) a- All CNVs span or break the NRXN1 gene; b-CNV appeared interesting in Wang et al. 2011; AN- anorexia nervosa; ctrl- controls; OR- odds ratio; CI- confidence interval.
176 Chapter 7
Power
With 2603 cases and 9101 controls, assuming CNV frequency of 0.04%, AN prevalence of 0.03%, setting alpha level at 0.05, we would be able to detect a CNV conferring a relative risk of about 3.8 with 80% probability. For reference, studies of SCZ show that a genetic risk of an associated rare CNV can range from 5 to 20 17. The alpha level of 0.05 adjusted for 20 tests (Bonferroni correction) is 0.0025. With alpha of 0.0025 and other assumptions remaining the same, we would be able to detect a CNV conferring a relative risk of about 5.25 with 80% probability.
Discussion
None of the candidate CNVs tested in the present study was significantly associated with AN. Although our sample size was sufficient to detect potent CNVs, very subtle effects might well remain undetected. The CNVs selected for this study included classical rare CNVs, which had previously been associated with SCZ, ASD or developmental delay, as well as candidates which had been known to associate with obesity-related syndromes. It should be acknowledged that even though we were able to utilize the largest AN sample size to date in ED genetic studies, it remains a relatively modest a sample size by contemporary standards compared to studies of rare CNVs in neurodevelopmental disorders such as SCZ. Rare CNVs associated with neuropsychiatric diseases are considered to be "potent", with effect sizes ranging from an OR of 5 to 20 55. Although the present study found no significant associations of the candidate CNVs in AN, some observations are noteworthy. A deletion of 15q13.3 could merit consideration for further investigation, as it was four times more frequent in cases than in controls. However, due to very low frequency, the significance of this observation cannot be determined in the present study.
177 Chapter 7
We also tested several rare CNVs which had been associated with obesity. Since the obesity phenotype in the CNV studies is often accompanied by developmental delay 19,56, it is not clear whether the effects of these CNVs on body weight are primary (by affecting homeostatic or hedonic food intake circuitry) or secondary (mediated by changes in behavior entailed by developmental delay, lower IQ, or psychiatric disorders). One interesting region is on chr16p11.2 where the deletion is associated with obesity 56, whereas duplication in the same locus has been associated with being underweight (defined as body mass index below 18.5 kg/m2) 18. These findings suggest a possible gene dosage effect on body weight on chr16p11.2. Since the cardinal symptom of AN is maintenance of very low body weight, Jacquemont et al.(2011) tested this same duplication in a sample of patients with eating disorders (including 109 patients with AN), but found no evidence of association. We too found no association, despite the larger sample size. It should be noted that we split the locus on 16p11.2 into two regions - proximal and distal - as in 41, but, there would clearly be no effect even if these regions had been considered jointly. Our results, in adjunction with the previously published reports, suggest that the etiology of AN may be distinct from being underweight or obese, at the level of both rare CNVs (as shown in the present study) and common SNPs 13. Wang et al. (2011) analysed some of the CNV regions covered in the present study in a group of 1033 cases with AN and 3773 controls (it should be noted that some of these controls overlap with the controls from the CHOP group used in the present study). None of the regions in question were associated with AN - a deletion at chr15q11.2 yielded a p-value of 0.036, but it did not survive adjustment for multiple testing. The possibility that insufficient statistical power was the reason for the lack of a significant association has been ruled out in the present study, which failed to demonstrate association with these CNVs in a larger sample size (OR of 1.05 and p-value of 0.85). It should be noted that Wang et al. (2011) investigated CNVs larger than 500kb only, whereas the present study included CNVs larger than 200kb).
178 Chapter 7
In addition to the candidate regions, Wang et al. also examined large and rare CNVs genome-wide in order to determine whether particular CNVs tended to cluster in the same genomic region. We tested two regions identified by Wang et al. 21 (on chr3p26.3 and chr13q12.12), neither of which showed evidence of association with AN.
A question arises whether PennCNV could have missed true calls (a false-negative error) in the detection phase or whether some samples with valid calls could have been excluded in the QC phase. This study was focused on large CNVs, which typically are called unambiguously, thus the sample- level QC thresholds (such as LogRR SD, BAF drift or the total number of CNVs per sample) were not extremely strict (see Supplementary Figures 1-4 for histograms showing the distribution of samples per each parameter's range). Thus, the post-detection QC was unlikely to lead to a false-negative error. Before entering the phase of running the detection algorithm, the cases and controls were subject to strict QC procedures, typical for large GWA studies (although they were not uniform across all groups). Whether the individuals with large CNVs might have a lower chance of passing the GWAS QC remains a question. If such a tendency is confirmed in future studies, it could theoretically lead to a bias via a false-negative error.
Limitations
The present results should be viewed in the context of several serious limitations. First, it was not possible to confirm the detected CNVs in vivo. In vivo confirmation via qPCR, MLPA or other techniques is often performed to replicate results and identify potential false positive CNV findings arising from imperfect CNV calling by the algorithm. Nonetheless, the present study focused on large CNVs (larger than 200kb), which contained many probes (the smallest number of probes in a CNV was 84). Calling of such large CNVs is unambiguous, and sensitivity and specificity remain very high 29. Additionally, confirmation of calls by two independent calling algorithms
179 Chapter 7
(PennCNV and QuantiSNP) and visual inspection of the graphs presenting LogRR and BAF per each CNV were performed to ensure that no false positives are included in the analysis. Another caveat is that although all the cases and controls in the study were of European descent, it was not possible to apply any dimensionality reduction technique to ethnically match cases and controls (raw genotype data were not available for most of the control groups). Control groups were subjects to different QC routines, which might be a source of potential systematic error. Furthermore, the cases and the controls were genotyped on four separate Illumina platforms. All CNV regions in question have dense SNP coverage on those platforms, but a platform-specific bias cannot be ruled out. Finally, although the cases were composed of females only, most of the controls were of mixed sex, and it has been shown that some of the candidate CNVs may have sex-specific effects 57. Importantly, the differences in the total CNV frequencies across the control groups do suggest a possible bias. All the limitations described above could be expected to increase chances of a false positive error rather than a false negative error. Unfortunately, both scenarios cannot be ruled out with the present study design and the data at hand (the intensity data were not available for the most of the control sample).
Conclusions
Whether AN is a neurodevelopmental disorder remains an open question. Some pieces of evidence suggest so 15, but the present study, which investigated a number of CNVs previously associated with neurodevelopmental conditions, could not support this claim. Our conclusions are limited to the CNVs in question and should be viewed in the context of serious limitations described above. These results do not rule out the possibility that there are other CNVs (not tested in the present study) or other forms of genetic variation which are associated with both AN and neurodevelopmental disorders.
180 Chapter 7
Acknowledgements
We are deeply thankful to all the participants and the control data providers.
Supplementary Material
The sample and SNP QC, as part of a GWAS of AN
Cases were controlled for the individual call rate (removal threshold at 99%), heterozygosity (subjects whose number of heterozygous genotypes was higher than 3 sd or lower than -3 sd were removed). The dataset of cases was merged with the HapMap populations for the multidimensional scaling analysis. The cases not matching the cluster of subjects with European ethnicity were removed. Subjects were also excluded on the basis of having incorrect sex assignments or evidence of cryptic relatedness (one of a cryptically related pair, as indicated by PI-HAT score > 0.05, was removed). SNPs with a call rate lower than 99% or the ones for which the Hardy-Weinberg equilibrium was violated at p<10-4 were removed. We also excluded SNPs with a minor allele frequency less than 1%. More details of this part of QC can be found in 1. Data were handled and analysed with PLINK, version 1.07 58.
Histograms from CNV QC
181 Chapter 7
Fig.1. Histogram. The Y axis represents number of samples per indicated bin (range), the X axis represents total number of CNVs per individual (genome- wide). Individuals with more than 300 CNVs were excluded.
Fig.2. Histogram. On Y axis the number of samples, on X axis standard deviation of the LogR Ratio. Individuals with LogRR SD larger than 0.35 were excluded.
Fig.3. Histogram. On Y axis the number of samples, on X axis BAF drift value (derived from BAF distribution). Individuals with BAF drift larger than 0.002 were excluded.
182 Chapter 7
Fig.4. Histogram. On Y axis the number of samples, on X axis waviness factor. Individuals with waviness factor larger than 0.05 or smaller than -0.05 were excluded.
Exploratory analyses
Three regions were excluded from the analysis at various quality control steps. We show the data pertaining to them here, as they might be of interest.
Chr17q11.2 The calls made by PennCNV in this region were not confirmed by QuantiSNP. Looking at PennCNV calls only, the deletions on chr17p11.2 occurred more often in cases than controls (OR=4.64 and unadjusted p-value=0.0006) (Supplementary Table 2). Supplementary Figures 5a-5d show example graphs of LogRR and BAF of 4 PennCNV calls.
183 Chapter 7
Chr1p21.3 The calls made by PennCNV in this region were not confirmed by QuantiSNP. Looking at PennCNV calls only, the deletions on chr1p21.3 occurred more often in cases than controls (OR=5.23 and unadjusted p-value=0,011) (Supplementary Table 2). Supplementary Figures 6a-6d show example graphs of LogRR and BAF of 4 PennCNV calls.
184 Chapter 7
Chr5q13.2 Another region which might be of interest for further study is at 5q13.2 (chr5:68,500,000-70,700,000). We were unable to properly test this region due to the lack of probe coverage in most of our control groups. Cases with AN and two small control groups (Finnish and Greek) were genotyped on Illumina 670 platform, which covers this region with monomorphic CNV- probes, whereas most of the older Illumina chips do not. PennCNV algorithm detected 8 duplications in 8 cases and 0 in controls (ncases=2603, ncontrols=475). These 8 duplications were also detected by QuantiSNP. Supplementary Figures 7a-7d show example graphs of LogRR and BAF of 4 duplications.
185 Chapter 7
Supplementary Table 2. Counts and frequencies of 3 regions which were not included in the main analysis due to either the lack of confirmation of calls by QuantiSNP or insufficient probe coverage in the control groups. Cases Ctrl Ctrl Ctrl Ctrl Ctrl Ctrl N => 2603 79 396 576 5089 1641 1320 2603 9101
Fisher's CNV Sta WTCC Twins Sum Sum OR (-95%- AN Greek Fin. Dutch CHOP exact p- region te C2 UK Cases Ctrl 95%CI) val
17p11 4,64 (1,82- Del 12 3 0 0 6 0 0 12 9 0,0006 .2a, c 10,3) freq. 0,46% 3,80% 0,00% 0,00% 0,12% 0,00% 0,00% 0,46% 0,10% 1p21. 5,23 Del 6 1 1 0 2 0 0 6 4 0,011 3c (1,37-17,3) freq. 0,23% 1,27% 0,25% 0,00% 0,04% 0,00% 0,00% 0,23% 0,04% 5q13. Du Inf 8 0 0 - - - - 8 0 0,62 2b p (0-inf) freq. 0,31% 0,00% 0,00% - - - - 0,31% 0,00%
186 Chapter 7
a- All CNVs span or break the RAI1 gene; b-data unavailable for 4 control groups due to insufficient probe coverage; c-calls by PennCNV not replicated by QuantiSNP; AN-anorexia nervosa; ctrl-controls; OR-odds ratio; CI-confidence interval; Del-deletion; Dup-duplication.
References
1. Boraska V, Franklin C, Floyd J, Thornton L, Huckins L, Southam L, et al. A genome-wide association study of anorexia nervosa. Mol Psychiatry 2014;19:1085-1094. 2. Birmingham CL, Su J, Hlynsky JA, Goldner EM, Gao M. The mortality rate from anorexia nervosa. Int J Eat Disord 2005;38:143-146. 3. Papadopoulos FC, Ekbom A, Brandt L, Ekselius L. Excess mortality, causes of death and prognostic factors in anorexia nervosa. The British Journal of Psychiatry 2009;194:10-17. 4. Kortegaard LS, Hoerder K, Joergensen J, Gillberg C, Kyvik KO. A preliminary population-based twin study of self-reported eating disorder. Psychol Med 2001;31:361-365. 5. Wade TD, Bulik CM, Neale M, Kendler KS. Anorexia nervosa and major depression: Shared genetic and environmental risk factors. Am J Psychiatry 2000;157:469-471. 6. Bulik CM, Thornton LM, Root TL, Pisetsky EM, Lichtenstein P, Pedersen NL. Understanding the relation between anorexia nervosa and bulimia nervosa in a swedish national twin sample. Biol Psychiatry 2010;67:71-77. 7. Strober M, Freeman R, Lampert C, Diamond J, Kaye W. Controlled family study of anorexia nervosa and bulimia nervosa: Evidence of shared liability and transmission of partial syndromes. Am J Psychiatry 2000;157:393-401. 8. Oldershaw A, Treasure J, Hambrook D, Tchanturia K, Schmidt U. Is anorexia nervosa a version of autism spectrum disorders? European Eating Disorders Review 2011;19:462-474. 9. Miotto P, Pollini B, Restaneo A, Favaretto G, Sisti D, Rocchi MBL, et al. Symptoms of psychosis in anorexia and bulimia nervosa. Psychiatry Res 2010;175:237-243.
187 Chapter 7
10. Thorpe M, Nance M, Gilchrist P, Schutz J. Symptoms of psychosis in a patient with anorexia nervosa. Aust N Z J Psychiatry 2011;45:791-791. 11. Altman SE, Shankman SA. What is the association between obsessive– compulsive disorder and eating disorders? Clin Psychol Rev 2009. 12. Collier DA, Treasure JL. The aetiology of eating disorders. The British Journal of Psychiatry 2004;185:363-365. 13. Brandys MK, van Elburg AA, Loos RJF, Bauer F, Hendriks J, van der Schouw YT, et al. Are recently identified genetic variants regulating BMI in the general population associated with anorexia nervosa? American Journal of Medical Genetics Part B: Neuropsychiatric Genetics 2010;153B:695-699. 14. Connan F, Campbell IC, Katzman M, Lightman SL, Treasure J. A neurodevelopmental model for anorexia nervosa. Physiol Behav 2003;79:13- 24. 15. Kerbeshian J, Burd L. Is anorexia nervosa a neuropsychiatric developmental disorder? an illustrative case report. World Journal of Biological Psychiatry 2009;10:648-657. 16. de Kovel CGF, Trucks H, Helbig I, Mefford HC, Baker C, Leu C, et al. Recurrent microdeletions at 15q11.2 and 16p13.11 predispose to idiopathic generalized epilepsies. Brain 2010;133:23-32. 17. Sullivan PF, Daly MJ, O'Donovan M. Genetic architectures of psychiatric disorders: The emerging picture and its implications. Nat Rev Genet 2012. 18. Jacquemont S, Reymond A, Zufferey F, Harewood L, Walters RG, Kutalik Z, et al. Mirror extreme BMI phenotypes associated with gene dosage at the chromosome 16p11. 2 locus. Nature 2011;478:97-102. 19. D'Angelo CS, Koiffmann CP. Copy number variants in obesity-related syndromes: Review and perspectives on novel molecular approaches. Journal of Obesity 2012;2012. 20. Helder SG, Collier DA. The genetics of eating disorders. Curr Top Behav Neurosci 2011;6:157-175. 21. Wang K, Zhang H, Bloss CS, Duvvuri V, Kaye W, Schork NJ, et al. A genome-wide association study on common SNPs and rare CNVs in anorexia nervosa. Mol Psychiatry 2011;16:949-959.
188 Chapter 7
22. Poyastro Pinheiro A, Thornton LM, Plotonicov KH, Tozzi F, Klump KL, Berrettini WH, et al. Patterns of menstrual disturbance in eating disorders. Int J Eat Disord 2007;40:424-434. 23. Tozzi F, Thornton LM, Klump KL, Fichter MM, Halmi KA, Kaplan AS, et al. Symptom fluctuation in eating disorders: Correlates of diagnostic crossover. Am J Psychiatry 2005;162:732-740. 24. Anderson CA, Pettersson FH, Clarke GM, Cardon LR, Morris AP, Zondervan KT. Data quality control in genetic case-control association studies. Nature protocols 2010;5:1564-1573. 25. Rucker JJH, Breen G, Pinto D, Pedroso I, Lewis CM, Cohen-Woods S, et al. Genome-wide association analysis of copy number variation in recurrent depressive disorder. Mol Psychiatry 2011. 26. Spector TD, MacGregor AJ. The st. thomas' UK adult twin registry. Twin Research 2002;5:440-443. 27. Stefansson H, Ophoff RA, Steinberg S, Andreassen OA, Cichon S, Rujescu D, et al. Common variants conferring risk of schizophrenia. Nature 2009;460:744-747. 28. Shaikh TH, Gai X, Perin JC, Glessner JT, Xie H, Murphy K, et al. High- resolution mapping and analysis of copy number variations in the human genome: A data resource for clinical and research applications. Genome Res 2009;19:1682-1690. 29. Grozeva D, Conrad DF, Barnes CP, Hurles M, Owen MJ, O'Donovan MC, et al. Independent estimation of the frequency of rare CNVs in the UK population confirms their role in schizophrenia. Schizophr Res 2012. 30. Rosenfeld JA, Traylor RN, Schaefer GB, McPherson EW, Ballif BC, Klopocki E, et al. Proximal microdeletions and microduplications of 1q21. 1 contribute to variable abnormal phenotypes. European Journal of Human Genetics 2012;20:754-761. 31. Kirov G, Pocklington A, Holmans P, Ivanov D, Ikeda M, Ruderfer D, et al. De novo CNV analysis implicates specific abnormalities of postsynaptic signalling complexes in the pathogenesis of schizophrenia. Mol Psychiatry 2011;17:142-153.
189 Chapter 7
32. Levinson DF, Duan J, Oh S, Wang K, Sanders AR, Shi J, et al. Copy number variants in schizophrenia: Confirmation of five previous findings and new evidence for 3q29 microdeletions and VIPR2 duplications. Am J Psychiatry 2011;168:302-316. 33. Crespi BJ, Crofts HJ. Association testing of copy number variants in schizophrenia and autism spectrum disorders. Journal of Neurodevelopmental Disorders 2012;4:15. 34. Stefansson H, Rujescu D, Cichon S, Pietilainen OPH, Ingason A, Steinberg S, et al. Large recurrent microdeletions associated with schizophrenia. Nature 2008;455:232-236. 35. Malhotra D, McCarthy S, Michaelson JJ, Vacic V, Burdick KE, Yoon S, et al. High frequencies of de novo CNVs in bipolar disorder and schizophrenia. Neuron 2011;72:951-963. 36. Sanders SJ, Ercan-Sencicek AG, Hus V, Luo R, Murtha MT, Moreno-De- Luca D, et al. Multiple recurrent de novo CNVs, including duplications of the 7q11. 23 williams syndrome region, are strongly associated with autism. Neuron 2011;70:863-885. 37. Kirov G, Rujescu D, Ingason A, Collier DA, O'Donovan MC, Owen MJ. Neurexin 1 (NRXN1) deletions in schizophrenia. Schizophrenia Bulletin 2009. 38. Williams NM, Franke B, Mick E, Anney RJL, Freitag CM, Gill M, et al. Genome-wide analysis of copy number variants in attention deficit hyperactivity disorder: The role of rare variants and duplications at 15q13. 3. Am J Psychiatry 2012;169:195-204. 39. Williams NM, Zaharieva I, Martin A, Langley K, Mantripragada K, Fossdal R, et al. Rare chromosomal deletions and duplications in attention-deficit hyperactivity disorder: A genome-wide analysis. Lancet 2010. 40. Bachmann-Gagescu R, Mefford HC, Cowan C, Glew GM, Hing AV, Wallace S, et al. Recurrent 200-kb deletions of 16p11. 2 that include the SH2B1 gene are associated with developmental delay and obesity. Genetics in Medicine 2010;12:641-647. 41. Guha S, Rees E, Darvasi A,et al. IMplication of a rare deletion at distal 16p11.2 in schizophrenia. JAMA Psychiatry 2013:1-8.
190 Chapter 7
42. McCarthy SE, Makarov V, Kirov G, Addington AM, McClellan J, Yoon S, et al. Microduplications of 16p11.2 are associated with schizophrenia. Nat Genet 2009;41:1223-1227. 43. Kirov G, Grozeva D, Norton N, Ivanov D, Mantripragada KK, Holmans P, et al. Support for the involvement of large copy number variants in the pathogenesis of schizophrenia. Hum Mol Genet 2009;18:1497-1503. 44. Vieira GH, Rodriguez JD, Carmona-Mora P, Cao L, Gamba BF, Carvalho DR, et al. Detection of classical 17p11. 2 deletions, an atypical deletion and RAI1 alterations in patients with features suggestive of Smith–Magenis syndrome. European Journal of Human Genetics 2011. 45. Carmona-Mora P, Walz K. Retinoic acid induced 1, RAI1: A dosage sensitive gene related to neurobehavioral alterations including autistic behavior. Curr Genomics 2010;11:607. 46. Carter M, Nikkel S, Fernandez B, Marshall C, Noor A, Lionel A, et al. Hemizygous deletions on chromosome 1p21. 3 involving the DPYD gene in individuals with autism spectrum disorder. Clin Genet 2011;80:435-443. 47. Willemsen MH, Vallès A, Kirkels LAMH, Mastebroek M, Loohuis NO, Kos A, et al. Chromosome 1p21. 3 microdeletions comprising DPYD and MIR137 are associated with intellectual disability. J Med Genet 2011;48:810-818. 48. Stevens SJC, van Ravenswaaij-Arts C, Janssen JWH, Klein Wassink-Ruiter JS, van Essen AJ, Dijkhuizen T, et al. MYT1L is a candidate gene for intellectual disability in patients with 2p25. 3 (2pter) deletions. American Journal of Medical Genetics Part A 2011. 49. Lee Y, Mattai A, Long R, Rapoport JL, Gogtay N, Addington AM. Microduplications disrupting the MYT1L gene (2p25. 3) are associated with schizophrenia. Psychiatr Genet 2012;22:206-209. 50. Shinawi M, Sahoo T, Maranda B, Skinner S, Skinner C, Chinault C, et al. 11p14. 1 microdeletions associated with ADHD, autism, developmental delay, and obesity. American Journal of Medical Genetics Part A 2011;155:1272-1280.
191 Chapter 7
51. Diskin SJ, Li M, Hou C, Yang S, Glessner J, Hakonarson H, et al. Adjustment of genomic waves in signal intensities from whole-genome SNP genotyping platforms. Nucleic Acids Res 2008;36:e126-e126. 52. Wang K, Li M, Hadley D, Liu R, Glessner J, Grant SF, et al. PennCNV: An integrated hidden markov model designed for high-resolution copy number variation detection in whole-genome SNP genotyping data. Genome Res 2007;17:1665-1674. 53. Colella S, Yau C, Taylor JM, Mirza G, Butler H, Clouston P, et al. QuantiSNP: An objective bayes hidden-markov model to detect and accurately map copy number variation using SNP genotyping data. Nucl Acids Res 2007;35:2013-2025. 54. Purcell S, Cherny SS, Sham PC. Genetic power calculator: Design of linkage and association genetic mapping studies of complex traits. Bioinformatics 2003;19:149-150. 55. Collins AL, Sullivan PF. Genome-wide association studies in psychiatry: What have we learned? The British Journal of Psychiatry 2013;202:1-4. 56. Walters RG, Jacquemont S, Valsesia A, de Smith AJ, Martinet D, Andersson J, et al. A new highly penetrant form of obesity due to deletions on chromosome 16p11.2. Nature 2010;463:671-675. 57. Tropeano M, Ahn JW, Dobson RJ, Breen G, Rucker J, Dixit A, et al. Male- biased autosomal effect of 16p13. 11 copy number variation in neurodevelopmental disorders. PloS one 2013;8:e61365. 58. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MAR, Bender D, et al. PLINK: A tool set for whole-genome association and population-based linkage analyses. The American Journal of Human Genetics 2007;81:559-575. 59. Hoek HW. Incidence, prevalence and mortality of anorexia nervosa and other eating disorders. Curr Opin Psychiatry 2006;19:389-394.
192 Appendix
Appendix
A genome-wide association study of anorexia nervosa
Vesna Boraska Christopher S Franklin James AB Floyd Laura M Thornton Laura M Huckins Lorraine Southam N William Rayner Ioanna Tachmazidou Kelly L Klump Janet Treasure Cathryn M Lewis Ulrike Schmidt Federica Tozzi Kirsty Kiezebrink Johannes Hebebrand Philip Gorwood Roger AH Adan Martien JH Kas Angela Favaro Paolo Santonastaso Fernando Fernández-Aranda Monica Gratacos Filip Rybakowski Monika Dmitrzak-Weglarz Jaakko Kaprio Anna Keski-Rahkonen Anu Raevuori
193 Appendix
Eric F Van Furth Margarita CT Slof-Op t Landt James I Hudson Ted Reichborn-Kjennerud Gun Peggy S Knudsen Palmiero Monteleone Allan S Kaplan Andreas Karwautz Hakon Hakonarson Wade H Berrettini Yiran Guo Dong Li Nicholas J. Schork Gen Komaki Tetsuya Ando Hidetoshi Inoko Tõnu Esko Krista Fischer Katrin Männik Andres Metspalu Jessica H Baker Roger D Cone Jennifer Dackor Janiece E DeSocio Christopher E Hilliard Julie K O’Toole Jacques Pantel Jin P Szatkiewicz Chrysecolla Taico Stephanie Zerwas Sara E Trace Oliver SP Davis
194 Appendix
Sietske Helder Katharina Bühren Roland Burghardt Martina de Zwaan Karin Egberts Stefan Ehrlich Beate Herpertz-Dahlmann Wolfgang Herzog Hartmut Imgart André Scherag Susann Scherag Stephan Zipfel Claudette Boni Nicolas Ramoz Audrey Versini Marek K Brandys Unna N Danner Carolien de Kovel Judith Hendriks Bobby PC Koeleman Roel A Ophoff Eric Strengman Annemarie A van Elburg Alice Bruson Maurizio Clementi Daniela Degortes Monica Forzan Elena Tenconi Elisa Docamp Geòrgia Escaramís Susana Jiménez-Murcia Jolanta Lissowska
195 Appendix
Andrzej Rajewski Neonila Szeszenia-Dabrowska Agnieszka Slopien Joanna Hauser Leila Karhunen Ingrid Meulenbelt P Eline Slagboom Alfonso Tortorella Mario Maj George Dedoussis Dimitris Dikeos Fragiskos Gonidakis Konstantinos Tziouvas Artemis Tsitsika Hana Papezova Lenka Slachtova Debora Martaskova James L. Kennedy Robert D. Levitan Zeynep Yilmaz Julia Huemer Doris Koubek Elisabeth Merl Gudrun Wagner Paul Lichtenstein Gerome Breen Sarah Cohen-Woods Anne Farmer Peter McGuffin Sven Cichon Ina Giegling Stefan Herms
196 Appendix
Dan Rujescu Stefan Schreiber H-Erich Wichmann Christian Dina Rob Sladek Giovanni Gambaro Nicole Soranzo Antonio Julia Sara Marsal Raquel Rabionet Valerie Gaborieau Danielle M Dick Aarno Palotie Samuli Ripatti Elisabeth Widén Ole A Andreassen Thomas Espeseth Astri Lundervold Ivar Reinvang Vidar M Steen Stephanie Le Hellard Morten Mattingsdal Ioanna Ntalla Vladimir Bencko Lenka Foretova Vladimir Janout Marie Navratilova Steven Gallinger Dalila Pinto Stephen Scherer Harald Aschauer Laura Carlberg
197 Appendix
Alexandra Schosser Lars Alfredsson Bo Ding Lars Klareskog Leonid Padyukov Chris Finan Gursharan Kalsi Marion Roberts Darren W Logan Leena Peltonen Graham RS Ritchie Jeffrey C Barrett The Wellcome Trust Case Control Consortium 3 Xavier Estivill Anke Hinney Patrick F Sullivan David A Collier Eleftheria Zeggini Cynthia M Bulik
Molecular Psychiatry 2014; 19(10):1085-94.
My role in this paper included management of the database of the Dutch cases with AN, collection and storage of the phenotypic data, and selection and delivery of cases qualified for genotyping, according to GCAN criteria. Furthermore, as a member of the GCAN analysis team I was involved in the conceptual phase of the paper, searching for the case and control groups and I ran the quality control of the control groups during two two- week visits in the Sanger Institute.
198 Appendix
Abstract
Anorexia nervosa (AN) is a complex and heritable eating disorder characterized by dangerously low body weight. Neither candidate gene studies nor an initial genome wide association study (GWAS) have yielded significant and replicated results. We performed a GWAS in 2,907 cases with AN from 14 countries (15 sites) and 14,860 ancestrally matched controls as part of the Genetic Consortium for AN (GCAN) and the Wellcome Trust Case Control Consortium 3 (WTCCC3). Individual association analyses were conducted in each stratum and meta-analyzed across all 15 discovery datasets. Seventy-six (72 independent) SNPs were taken forward for in silico (two datasets) or de novo (13 datasets) replication genotyping in 2,677 independent AN cases and 8,629 European ancestry controls along with 458 AN cases and 421 controls from Japan. The final global meta-analysis across discovery and replication datasets comprised 5,551 AN cases and 21,080 controls. AN subtype analyses (1,606 AN restricting; 1,445 AN binge-purge) were performed. No findings reached genome-wide significance. Two intronic variants were suggestively associated: rs9839776 (P=3.01x10-7) in SOX2OT and rs17030795 (P=5.84x10-6) in PPP3CA. Two additional signals were specific to Europeans: rs1523921 (P=5.76x10-6) between CUL3 and FAM124B and rs1886797 (P=8.05x10-6) near SPATA13. Comparing discovery to replication results, 76% of the effects were in the same direction, an observation highly unlikely to be due to chance (P= 4x10-6), strongly suggesting that true findings exist but that our sample, the largest yet reported, was underpowered for their detection. The accrual of large genotyped AN case-control samples should be an immediate priority for the field.
199 Appendix
Introduction
Anorexia nervosa (AN) is a perplexing biologically-influenced psychiatric disorder characterized by the maintenance of dangerously low body weight, fear of weight gain, and seeming indifference to the seriousness of the illness.1 AN affects ~1% of the population.2, 3 Females are disproportionately afflicted, although males also develop the condition.4 The most common age of onset is 15-19 years;5 however, the incidence appears to be increasing in the pre-pubertal period6 and in older adults.7 AN is often comorbid with major depressive disorder, anxiety disorders, and multiple somatic complications.8-12 Although most individuals recover, ~25% develop a chronic and relapsing course.13 AN ranks among the ten leading causes of disability among young women14 and has one of the highest mortality rates of any psychiatric disorder.15-19 The evidence base for treatment for AN has been described as “weak,”20, 21 and treatment and extended inpatient hospitalizations for weight restoration are costly.22, 23 In sum, the public health impact of AN is considerable, and AN carries substantial morbidity, mortality, and personal, familial, and societal costs.
As with most idiopathic psychiatric disorders, the inheritance of AN is complex. The core features of AN [i.e., the ability and determination to maintain low body mass index (BMI)] are remarkably homogeneous across time and cultures.24, 25 Genetic epidemiological studies have documented the familiality of AN (relative risk 11.3 in first-degree relatives of AN probands)26, 27 and the estimated twin-based heritability of AN ranges from 33 to 84%.28-32 Genome-wide linkage studies did not narrow the genomic search space in a compelling manner.33-35 Findings from candidate gene studies of AN resemble those for most complex biomedical diseases—initial intriguing findings diminished by the absence of rigorous replication.36-38
200 Appendix
Given the centrality of weight dysregulation to AN, genes implicated in the regulation of body weight might also be involved in the etiology of AN.39, 40 Therefore genetic variants with a profound effect on BMI are worthy of consideration.38
Two genome-wide association studies (GWAS) of AN have been conducted. One study that used DNA pooling and genotyping with a modest number of microsatellite markers with follow-up genotyping detected evidence for association with rs2048332 on chromosome 1, but this finding did not reach genome-wide significance.41 A GWAS of 1033 AN cases from the USA, Canada, and Europe compared with 3733 pediatric controls yielded no genome-wide significant findings.42 Recently, a sequencing and genotyping study of 152 candidate genes in 1205 AN cases and 1948 controls suggested a novel association of a cholesterol metabolism influencing EPHX2 gene with susceptibility to AN.43
In recognition of the need for large-scale sample collections to empower GWAS, we established the Genetic Consortium for Anorexia Nervosa (GCAN) in 2007—a worldwide collaboration combining existing DNA samples of AN patients into a single resource. As part of the Wellcome Trust Case Control Consortium 3 (WTCCC3), we have conducted the largest GWAS for AN to date.
Materials and Methods
Discovery dataset. We conducted a GWAS across 15 discovery datasets, comprising a total of 2,907 AN cases and 14,860 ancestrally matched controls of European origin (Table 1). All AN cases were female. Diagnostic determination was via semi-structured or structured interview or population assessment strategy based on DSM diagnostic criteria. Cases met DSM-IV criteria for lifetime AN (restricting or binge-purge subtype) or lifetime DSM-
201 Appendix
IV eating disorders “not otherwise specified” (EDNOS) AN-subtype (i.e., exhibiting the core features of AN). We did not require the presence of amenorrhea as this criterion does not increase diagnostic specificity.44, 45 Given the frequency of diagnostic crossover, a lifetime history of bulimia nervosa was allowed.46 Exclusion criteria included the diagnosis of medical or psychiatric conditions that might have confounded the diagnosis of AN (e.g., psychotic disorders, mental retardation, or a medical or neurological condition causing weight loss). Controls were carefully selected to match for ancestry within each site and chosen primarily from existing GWAS genotypes through collaboration and genotyping repository (dbGAP) access. Each site obtained ethical approval from the local ethics committee, and all participants provided written informed consent in accordance with the Declaration of Helsinki.
Genotyping, imputation and quality control. AN cases from the 15 sites were genotyped using Illumina 660W-Quad arrays (Illumina, Inc., San Diego, CA, USA) at the Wellcome Trust Sanger Institute. Funding was available only for genotyping AN cases. Thus, control genotypes were selected from existing datasets matched as closely as possible to the ancestry of cases and Illumina arrays as similar as possible to the 660W array (Table S1). Quality control (QC) of directly typed variants was performed within each of the 15 case-control datasets (Table S2, Supplementary Information).
Phasing and imputation was performed separately for each of the 15 datasets using a common set of single nucleotide polymorphisms (SNPs) passing QC (Table S2) using the program Impute2 v2.1.2 (Supplementary Information).47 The imputation reference panel was HapMap 3 release 2. We used all available HapMap3 populations for imputation as it was shown that the increase in the reference panel decreases error.48, 49 Post-imputation filters were applied to remove SNPs with INFO scores < 0.4 or with MAF < 0.05. We observe high imputation accuracy (as captured by the INFO score) across a range of minor allele frequencies (Figure S1). There was high
202 Appendix concordance between directly genotyped variants with imputed dosages of the same variants after masking (Figure S2). Statistical analysis. Single-SNP association analyses were performed under an additive genetic model separately within each of the 15 datasets (Supplementary Information). We tested for association across the autosomes and the non-pseudoautosomal region of the X chromosome. Imputation and association analysis of the non-pseudoautosomal region of the chromosome X data were based on females (2,907 AN cases and 10,594 controls). Association analyses were performed using SNPTEST v2.2.049 under an additive model and using a score test. To guard against false positives due to population stratification, we carried out association analysis within each dataset and then combined the results using meta-analysis (for the French dataset, the first principal component was added as a covariate). Fixed- effects meta-analyses were performed using GWAMA.50 All 15 discovery datasets were corrected for the genomic control (GC) inflation factor (λGC) prior to performing meta-analysis (Table S2; Supplementary Information).
Replication. We prioritized directly genotyped and imputed SNPs for replication based on statistical significance (P < 10-4), robust QC metrics, and vicinity to plausible candidate genes. In total 96 SNPs (95 autosomal and one on chromosome X) in 66 genomic regions showed nominal evidence for association. We selected 72 independent, uncorrelated variants representing each of the 66 associated genomic regions and added 4 proxies for the most associated SNPs resulting in 76 SNPs for replication. Cluster plots of all prioritized SNPs were examined using Evoker51 in cases and controls separately to minimize the possibility of spurious association due to genotyping error. We included 27 ancestry-informative markers (AIMs) for genotyping in the replication datasets, to guard against population stratification (Supplementary Information).52
Our replication data included 15 datasets—two existing in silico datasets and 13 datasets for de novo genotyping (Table 1). The in silico
203 Appendix dataset from the USA came from an existing GWAS of AN genotyped using the Illumina HumanHap610 platform (Illumina, San Diego, CA, USA)53 and the other in silico dataset came from Estonian Genome Center (www.biobank.ee) and was genotyped using the Illumina OmniExpress array. De novo genotyped samples included newly collected AN cases and controls from members of the GCAN and samples from the same sites as the discovery samples that had failed GWAS QC (including saliva and whole genome amplified samples). De novo SNP genotyping was carried out using the iPLEX Gold Assay (Sequenom, Inc., San Diego, CA, USA). SNPs with poor Sequenom design metrics were replaced with high-LD proxies. Sample and SNP QC were performed within each replication dataset. QC included checking for sex inconsistencies and exclusions based on sample call rate <80%, SNP call rate <90% and exact Hardy-Weinberg Equilibrium (HWE) p<0.0001. In total, replication genotypes (in silico and de novo) of 76 prioritized SNPs and 27 AIMs were available from 2,677 AN cases and 8,629 controls of European ethnicity and 458 AN cases and 421 controls from Japan.
Association analyses of prioritized SNPs were performed under an additive genetic model within each replication dataset with and without adjustment for AIMs. AIMs that showed nominally significant p-values for allele frequency differences between de novo typed cases and controls were used for conditional analysis (Table S3). As there were no qualitative differences between these results, the main text reports the unadjusted results. The USA replication dataset contained individuals who were related to individuals from the USA discovery dataset. As such, those samples were excluded from the discovery dataset and combined with replication USA samples to correctly account for relatedness between samples for the final global meta-analysis and sign test. Software packages GenABEL54 and GEMMA55 were used for replication analysis of the USA dataset. Fixed-effects meta-analysis across the replication datasets was performed using GWAMA50 (with and without adjustment for AIMs and in samples of European ancestry
204 Appendix only, i.e., excluding Japan, also with and without adjustment for AIMs). We also performed meta-analyses across the discovery and replication datasets, comprising a total of 5,551 AN cases and 21,080 controls (USA discovery samples were included only once as part of the replication phase). We calculated the power of the final global meta-analysis using QUANTO.56
Seventy-two independent SNPs were used to compare the direction of effects between the discovery and replication meta-analyses using R.57 For this analysis, the USA samples were used only once as part of the replication meta-analysis.
Additional analyses. We performed three additional analyses: 1) genome- wide complex trait analysis (GCTA), designed to estimate the proportion of phenotypic variance explained by genome-wide SNPs for complex traits,58 a network analysis, and a gene-based association test (Supplementary Information).
AN subtype analyses. Two subtype (Supplementary Information) association analyses were performed for the 76 prioritized SNPs across the discovery and replication datasets (Table 1). In total, the AN restricting subtype global meta-analysis included 1,606 cases and the AN binge-purge subtype analysis included 1,445 cases. Both analyses used the same set of 16,303 controls (Supplementary Information).
Related traits. Using the discovery meta-analysis, we investigated evidence for association using SNP results from published studies: 9 SNPs with nominal evidence of association with AN;42 14 SNPs suggestively associated with eating disorder-related symptoms, behaviors, or personality traits;59, 60 89 SNPs with genome-wide significance in studies of BMI or obesity;61, 62 and 15 SNPs related to morbid obesity.61 We also investigated evidence for association across the 72 replication SNPs using published GWAS results from the Psychiatric Genomics Consortium (https://pgc.unc.edu) for
205 Appendix attention-deficit/ hyperactivity disorder (ADHD), schizophrenia, bipolar disorder, and major depressive disorder.63-66
Expression studies. We prioritized the top 20 SNPs in terms of statistical significance and quantified the expression of the two nearest genes per SNP (Table S4) in 12 inbred strains of mice. We obtained publicly available RNAseq data from whole brain tissue samples and used standard software to map and count the sequence reads (Supplementary Information).
Results
Main association results. Of 1,185,559 imputed SNPs that passed QC, 287 showed evidence for association in the discovery stage with P < 10-4. These variants represented 66 independent signals and had frequencies and effect sizes commensurate with observations in other common complex diseases. One variant, not surrounded by other SNPs achieving low p-values and for which genotypes were only available in two of the 15 initial study groups, surpassed genome-wide significance (rs4957798, P=1.67x10-12) but was not subsequently replicated in the global meta-analysis across discovery and replication samples. The overall λGC was 1.03 (Figures S3 and S4). Seventy-six SNPs (of which 72 were independent) were prioritized for follow-up through in silico and de novo replication (Table S5). Nine SNPs showed association with P < 0.05 (minimum p-value was 0.003) in the replication dataset meta- analysis (binomial P=0.0135) (Table S5). Based on 72 independent SNPs taken forward, we would expect 0.05x72=3.6 SNPs to reach P=0.05 by chance. The 0.0135 P value reflects this enrichment in signal. No signals surpassed genome-wide significance (P=5x10-8) in the final global meta- analysis across all discovery and replication samples (Table S5) or in the AN subtype analyses (Tables S6-S7).
206 Appendix
Of critical importance, we observed significant evidence of SNP effect sizes in the replication data in the same direction as the discovery set (55/72 signals, sign test binomial P=4x10-6). This enrichment was also observed for the AN restricting (58/72, P=8x10-8) and binge-purge (56/72, P=1x10-6) subtype analyses. These results strongly indicate that the prioritized set of variants is likely to contain true positive signals for AN but that the current sample size is insufficient to detect these effects.
Our analysis revealed two notable variants: rs9839776 (P=3.01x10-7) in SOX2OT (SOX2 overlapping transcript) and rs17030795 (P=5.84x10-6) in PPP3CA (protein phosphatase 3, catalytic subunit, alpha isozyme) (Table 2). Two additional signals emerged from the analysis focused on European replication samples only: rs1523921 (P=5.76x10-6) located between CUL3 (cullin 3) and FAM124B (family with sequence similarity 124B) and rs1886797 (P=8.05x10-6) located 18kb from SPATA13 (spermatogenesis associated 13) (Table S5). Four signals were in neurodevelopmental genes regulating synapse and neuronal network formation (SYN2, NCAM2, CNTNAP2 and CTNNA2; Table 2).
AN subtype analyses. In the AN restricting subtype analyses, the two most significant signals were rs1523921 (as in the main analysis, P=8.39x10-5) and rs10777211 (P=8.95x10-5) located 333kb from ATP2B1 (ATPase, calcium transporting, plasma membrane 1), both detected in the European-only analysis (Table S6). The most significant result for AN binge-purge analysis was rs9839776 (as in the main analysis, P=3.97x10-4) in SOX2OT, also in Europeans only (Table S7). Overall, signals from the main AN case-control analysis display similar levels of association across both AN subtypes (Table S8).
Additional analyses. GCTA is technically challenging when synthesizing data across multiple strata with small individual sample sizes. When we applied it to our data we saw great variability in the estimates of variance and did not
207 Appendix judge the results reliable. Results of the gene-based association test and network analysis are presented in their entirety in Supplemental Information and Figure S5, both of which were unremarkable.
Related traits. Nine out of the 11 previously reported variants suggestively associated with AN42 were found in our discovery meta-analysis, and six of these 9 SNPs had the same direction of effect as originally reported (P=0.508) (Table S9). Twelve out of 14 variants previously reported to be associated with eating disorder-related symptoms, behaviors, and personality traits59, 60 were found in our discovery meta-analysis and 7 had the same direction of effect (P=0.774) (Table S10), with one SNP (inside RUFY1) having P<0.05 (binomial P=0.459). We did not find evidence for signal enrichment in the 60 independent SNPs found in the Psychiatric Genomics Consortium data for ADHD, schizophrenia, bipolar disorder, or major depressive disorder63-66 (Table S11).
When we compared 76 (53 independent) SNPs from the AN results with 89 established BMI/obesity SNPs,61, 62 five SNPs (inside NEGR1, PTBP2, TMEM18, FTO and MC4R) had P<0.05 (binomial P=0.1906). Twenty-six of these 53 SNPs had the same direction of effect as originally reported (binomial P value=1) (Table S12). Thirteen of 15 SNPs associated with extreme obesity were extracted from our dataset and 9 of these were independent. Four of these 9 SNPs had the same direction of effect as originally reported (binomial P value=1) (Table S13). Three SNPs (in TMEM18, FTO and MC4R) had P<0.05 (binomial P value=0.0084), indicating modest enrichment of nominally associated SNPs from extreme obesity in our discovery dataset.
Expression studies. We analyzed RNAseq data for whole-brain tissue obtained from 12 different mouse strains (Figure 1). We performed this analysis for 32 mouse orthologues of the 34 human genes identified (Table S4). All 32 genes were expressed in the brain, above an average of 2 FPKM
208 Appendix
(Fragments Per Kilobase of exon per Million fragments mapped). Specifically, we find extremely high expression levels for Ppp3ca (FPKM value 36.40). Further, we find high expression for Sox2ot, with an FPKM value of 8.02, and similar expression values for Cul3 (10.01) and Ctnna2 (10.79).
Discussion
Given that the evidence base for the treatment of AN remains weak and that no effective medications for its treatment exist,20, 67 advances in our understanding of the underlying biology of the disorder are essential in order to develop novel therapeutics and to reduce the loss of life and diminution of quality of life associated with the disorder. The GCAN/WTCCC3 investigation represents an unprecedented international genetic collaboration in the study of AN, which sets the foundation for further genetic studies.
Our final global meta-analysis had 80% power to detect SNPs with allele frequency of 0.35 and genotypic relative risk of 1.15 (α=5x10-8, additive model).68 The AN subtype meta-analysis had 80% power to detect SNPs with allele frequency of 0.35 and genotypic relative risk 1.27 for the AN restricting subtype and 1.28 for the AN binge-purge subtype. Given these limitations in power, our strongest indicator that larger sample sizes could detect genetic variants associated with AN was revealed in the sign tests. The strong and significant evidence for SNP effect sizes in the same direction between discovery and replication sets (P=4x10-6) clearly suggests that larger sample sizes could successfully identify variants associated with AN and with the AN subtypes potentially enabling differentiation on a genetic level between restricting and binge/purge subtypes.
Several genetic variants were suggestively associated with AN (P<10- 5) (Table 2). Two variants, rs9839776 in SOX2OT and rs17030795 in PPP3CA, were identified through analysis of all discovery and replication datasets.
209 Appendix
Two additional variants with P<10-5, rs1523921 located between CUL3 and FAM124B and rs1886797 located near SPATA13, were identified through analysis of individuals of European descent only (Table S5), suggesting either heterogeneity in the effects of these SNPs by ancestry or low power. The genes displayed in Table 2 are discussed in greater detail in the Supplementary Information; however, we highlight that four of these variants are neurodevelopmental genes that regulate synapse and neuronal network formation (SYN2, NCAM2, CNTNAP2 and CTNNA2) and two have been associated with Alzheimer’s disease (SOX2OT and PPP3CA). Additionally, one of our prioritized SNPs (rs6558000) (Table S5) is located in close vicinity (9kb upstream) of the EPHX2 gene that was recently identified as a susceptibility locus to AN through candidate gene sequencing study of early-onset severe AN cases and controls.43
Our expression studies further extend the GWAS findings. It is reasonable, although perhaps not essential, to expect that genes implicated in AN be expressed in the brain. Supporting this assumption, 32 mouse orthologues of 34 human genes identified as being of interest were expressed at least at a low level in mouse brain. Moreover, genes corresponding to the more strongly associated genetic variants tended to be more highly expressed. For example, high FPKM values for Ppp3ca, Cul3, and Sox2ot underscore that these genes may play a neuropsychiatric role.
AN subtype analyses were included to determine whether differences might exist between the classic restricting subtype of AN and the subtype marked by dysregulation characterized by binge eating and/or purging behavior. These analyses had lower power due to the smaller sample sizes. Only two SNPs, rs1523921 (also found to be suggestively associated in the main case-control analysis) and rs10777211 located 333kb upstream of ATP2B1, showed association at the 10-5 significance level (Table S6). Similarly, subsequent analyses pertaining to associated phenotypes (weight regulation: BMI/obesity loci,40, 61, 69, 70 and loci for extreme obesity;61, 71, 72
210 Appendix psychiatric comorbidities: ADHD, schizophrenia, bipolar disorder, and major depressive disorder) or previous equivocal association findings for AN or eating disorders (AN variants,42 eating disorder related symptoms, behaviors, and personality traits variants59, 60) did not reveal significant findings. More adequately powered analyses that could allow us to detect variants that can distinguish between these two subtypes could be clinically meaningful in predicting clinical course and outcome and eventually in designing targeted therapeutics.
Our understanding of the fundamental genetic architectures of complex medical diseases and psychiatric disorders has expanded rapidly.73 It has also become manifestly clear that genomic searches for common variation via GWAS can successfully uncover biological pathways of etiological relevance. The major limitation to discovery is sample size.74 A recent GWAS for schizophrenia reported the identification of 22 genome- wide significant loci for schizophrenia (21,000 cases and 38,000 controls), and the results yielded multiple themes of clear biological and translational significance (e.g., calcium biology and miR-137 regulation).75 Moreover, given that cases and controls were derived from multiple sources and genotyped on multiple platforms, imputation was essential. Although effective, the preferred approach will always be to have samples genotyped on the same platform to maximize comparability and the capacity to identify genomic associations.
Although the underlying biology of AN remains incompletely understood, the relative homogeneity of the phenotype, replicated heritability estimates, and encouraging results of the sign tests presented herein strongly encourage continuing this path of discovery. Phenotypic refinement and the identification of biomarkers of illness (independent of biomarkers of starvation) could assist with identification of risk loci. We believe that the surest and fastest path to fundamental etiological knowledge about the biological basis of AN is via GWAS in larger samples.74
211 Appendix
This path is notably safe given that it relies on off-the-shelf technology whose utility has been proven in empirical results for multiple biomedical and psychiatric disorders. This approach is cost-effective due to recent sharp decreases in genotyping pricing. Therefore, we believe that accrual of large genotyped AN case-control samples should be an immediate priority for the field.
Acknowledgments
Please see the online version of the article for the full list of acknowledgements.
Conflicts of Interest
Patrick F. Sullivan was on the SAB of Expression Analysis (Durham, NC). Cynthia Bulik was a consultant for Shire Pharmaceuticals at the time the manuscript was written. Federica Tozzi was full time employee of GSK at the time when the study was performed. David A. Collier was employed by Eli Lilly UK for a portion of the time that this study was performed. James L. Kennedy has received honoraria from Eli Lilly and Roche. Robert D. Levitan has received honorarium from Astra-Zeneca.
212 Appendix
Supplementary information
Supplementary information is available online at http://www.nature.com/mp/journal/v19/n10/full/mp2013187a.html or http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4325090/
References
1. Klump KL, Bulik CM, Kaye WH, Treasure J, Tyson E. Academy for Eating Disorders position paper: eating disorders are serious mental illnesses. Int J Eat Disord 2009; 42: 97-103. 2. Hudson JI, Hiripi E, Pope HG, Jr., Kessler RC. The prevalence and correlates of eating disorders in the National Comorbidity Survey Replication. Biol Psychiatry 2007; 61: 348-358. 3. Hoek H, van Hoeken D. Review of the prevalence and incidence of eating disorders. Int J Eat Disord 2003; 34: 383-396. 4. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. Fourth Edition Text Revision. American Psychiatric Press: Washington, DC, 2000. 5. Lucas AR, Beard CM, O'Fallon WM, Kurland LT. 50-year trends in the incidence of anorexia nervosa in Rochester, Minn.: a population- based study. Am J Psychiatry 1991; 148: 917-922. 6. Nicholls DE, Lynn R, Viner RM. Childhood eating disorders: British National Surveillance Study. Br J Psychiatry 2011; 198: 295-301. 7. Bueno B, Krug I, Bulik C, Jiménez-Murcia S, Granero R, Thornton L, Penelo E, Menchón J, Sánchez I, Tinahones F, Fernández-Aranda F. Late onset eating disorders in Spain: Clinical characteristics and therapeutic implications. J Clin Psychol in press. 8. Katzman D. Medical complications in adolescents with anorexia nervosa: a review of the literature. Int J Eat Disord 2005; 37: S52-59.
213 Appendix
9. Sharp C, Freeman C. The medical complications of anorexia nervosa. Br J Psychiatry 1993; 162: 452-462. 10. Kaye W, Bulik C, Thornton L, Barbarich BS, Masters K, Group PFC. Comorbidity of anxiety disorders with anorexia and bulimia nervosa. Am J Psychiatry 2004; 161: 2215-2221. 11. Godart N, Flament M, Perdereau F, Jeammet P. Comorbidity between eating disorders and anxiety disorders: A review. Int J Eat Disord 2002; 32: 253-270. 12. Fernandez-Aranda F, Pinheiro AP, Tozzi F, Thornton LM, Fichter MM, Halmi KA, Kaplan AS, Klump KL, Strober M, Woodside DB, Crow S, Mitchell J, Rotondo A, Keel P, Plotnicov KH, Berrettini WH, Kaye WH, Crawford SF, Johnson C, Brandt H, La Via M, Bulik CM. Symptom profile of major depressive disorder in women with eating disorders. Aust N Z J Psychiatry 2007; 41: 24-31. 13. Berkman ND, Lohr KN, Bulik CM. Outcomes of eating disorders: a systematic review of the literature. Int J Eat Disord 2007; 40: 293- 309. 14. Mathers CD, Vos ET, Stevenson CE, Begg SJ. The Australian Burden of Disease Study: measuring the loss of health from diseases, injuries and risk factors. Med J Aust 2000; 172: 592-596. 15. Sullivan PF. Mortality in anorexia nervosa. Am J Psychiatry 1995; 152: 1073-1074. 16. Arcelus J, Mitchell AJ, Wales J, Nielsen S. Mortality rates in patients with anorexia nervosa and other eating disorders. A meta-analysis of 36 studies. Arch Gen Psychiatry 2011; 68: 724-731. 17. Birmingham C, Su J, Hlynsky J, Goldner E, Gao M. The mortality rate from anorexia nervosa. Int J Eat Disord 2005; 38: 143-146. 18. Millar HR, Wardell F, Vyvyan JP, Naji SA, Prescott GJ, Eagles JM. Anorexia nervosa mortality in Northeast Scotland, 1965-1999. Am J Psychiatry 2005; 162: 753-757.
214 Appendix
19. Zipfel S, Lowe B, Reas DL, Deter HC, Herzog W. Long-term prognosis in anorexia nervosa: lessons from a 21-year follow- up study. Lancet 2000; 355: 721-722. 20. Bulik CM, Berkman ND, Brownley KA, Sedway JA, Lohr KN. Anorexia nervosa treatment: a systematic review of randomized controlled trials. Int J Eat Disord 2007; 40: 310-320. 21. Eating disorders: Core interventions in the treatment and management of anorexia nervosa, bulimia nervosa and related eating disorders. http://www.nice.org.uk/page.aspx?o=101239. 2004, Accessed November 15, 2013. 22. McKenzie JM, Joyce PR. Hospitalization for anorexia nervosa. Int J Eat Disord 1992; 11: 235-241. 23. Krauth C, Buser K, Vogel H. How high are the costs of eating disorders - anorexia nervosa and bulimia nervosa - for German society? Eur J Health Econ 2002; 3: 244-250. 24. Striegel-Moore RH, Bulik CM. Risk factors for eating disorders. Am Psychol 2007; 62: 181-198. 25. Kas M, Kaye W, Mathes W, Bulik C. Interspecies genetics of eating disorders traits. Am J Med Genets Part B; Neuropsychiatr Genet 2009; 150B: 318-327. 26. Strober M, Freeman R, Lampert C, Diamond J, Kaye W. Controlled family study of anorexia nervosa and bulimia nervosa: evidence of shared liability and transmission of partial syndromes. Am J Psychiatry 2000; 157: 393-401. 27. Lilenfeld L, Kaye W, Greeno C, Merikangas K, Plotnikov K, Pollice C, Rao R, Strober M, Bulik C, Nagy L. A controlled family study of restricting anorexia and bulimia nervosa: comorbidity in probands and disorders in first-degree relatives. Arch Gen Psychiatry 1998; 55: 603-610. 28. Bulik C, Sullivan P, Tozzi F, Furberg H, Lichtenstein P, Pedersen N. Prevalence, heritability and prospective risk factors for anorexia nervosa. Arch Gen Psychiatry 2006; 63: 305-312.
215 Appendix
29. Klump KL, Miller KB, Keel PK, McGue M, Iacono WG. Genetic and environmental influences on anorexia nervosa syndromes in a population-based twin sample. Psychol Med 2001; 31: 737-740. 30. Wade TD, Bulik CM, Neale M, Kendler KS. Anorexia nervosa and major depression: shared genetic and environmental risk factors. Am J Psychiatry 2000; 157: 469-471. 31. Kortegaard LS, Hoerder K, Joergensen J, Gillberg C, Kyvik KO. A preliminary population-based twin study of self-reported eating disorder. Psychol Med 2001; 31: 361-365. 32. Bulik CM, Thornton LM, Root TL, Pisetsky EM, Lichtenstein P, Pedersen NL. Understanding the relation between anorexia nervosa and bulimia nervosa in a Swedish national twin sample. Biol Psychiatry 2010; 67: 71-77. 33. Grice DE, Halmi KA, Fichter MM, Strober M, Woodside DB, Treasure JT, Kaplan AS, Magistretti PJ, Goldman D, Bulik CM, Kaye WH, Berrettini WH. Evidence for a susceptibility gene for anorexia nervosa on chromosome 1. Am J Hum Genet 2002; 70: 787-792. 34. Devlin B, Bacanu S, Klump K, Bulik C, Fichter M, Halmi K, Kaplan A, Strober M, Treasure J, Woodside DB, Berrettini WH, Kaye WH. Linkage analysis of anorexia nervosa incorporating behavioral covariates. Hum Mol Genet 2002; 11: 689-696. 35. Bacanu S, Bulik C, Klump K, Fichter M, Halmi K, Keel P, Kaplan A, Mitchell J, Rotondo A, Strober M, Treasure J, Woodside D, Bergen A, Berrettini W, Kaye W, Devlin B. Linkage analysis of anorexia and bulimia nervosa cohorts using selected behavioral phenotypes as quantitative traits or covariates. Am J Med Genet B Neuropsychiatr Genet 2005; 139: 61-68. 36. Slof-Op 't Landt M, van Furth E, Meulenbelt I, Slagboom P, Bartels M, Boomsma D, Bulik C. Eating disorders: From twin studies to candidate genes and beyond. Twin Res Hum Genet 2005; 16: 467- 482.
216 Appendix
37. Bulik CM, Slof-Op't Landt MC, van Furth EF, Sullivan PF. The genetics of anorexia nervosa. Ann Rev Nutr 2007; 27: 263-275. 38. Hinney A, Scherag S, Hebebrand J. Genetic findings in anorexia and bulimia nervosa. Prog Mol Biol Transl Sci 2010; 94: 241-270. 39. Hebebrand J, Remschmidt H. Anorexia nervosa viewed as an extreme weight condition: genetic implications. Hum Genet 1995; 95: 1-11. 40. Muller TD, Greene BH, Bellodi L, Cavallini MC, Cellini E, Di Bella D, Ehrlich S, Erzegovesi S, Estivill X, Fernandez-Aranda F, Fichter M, Fleischhaker C, Scherag S, Gratacos M, Grallert H, Herpertz- Dahlmann B, Herzog W, Illig T, Lehmkuhl U, Nacmias B, Ribases M, Ricca V, Schafer H, Scherag A, Sorbi S, Wichmann HE, Hebebrand J, Hinney A. Fat mass and obesity-associated gene (FTO) in eating disorders: evidence for association of the rs9939609 obesity risk allele with bulimia nervosa and anorexia nervosa. Obes Facts 2012; 5: 408-419. 41. Nakabayashi K, Komaki G, Tajima A, Ando T, Ishikawa M, Nomoto J, Hata K, Oka A, Inoko H, Sasazuki T, Shirasawa S. Identification of novel candidate loci for anorexia nervosa at 1q41 and 11q22 in Japanese by a genome-wide association analysis with microsatellite markers. J Hum Genet 2009; 54: 531-537. 42. Wang K, Zhang H, Bloss CS, Duvvuri V, Kaye W, Schork NJ, Berrettini W, Hakonarson H. A genome-wide association study on common SNPs and rare CNVs in anorexia nervosa. Mol Psychiatry 2011; 16: 949-959. 43. Scott-Van Zeeland AA, Bloss CS, Tewhey R, Bansal V, Torkamani A, Libiger O, Duvvuri V, Wineinger N, Galvez L, Darst BF, Smith EN, Carson A, Pham P, Phillips T, Villarasa N, Tisch R, Zhang G, Levy S, Murray S, Chen W, Srinivasan S, Berenson G, Brandt H, Crawford S, Crow S, Fichter MM, Halmi KA, Johnson C, Kaplan AS, La Via M et al. Evidence for the role of EPHX2 gene variants in anorexia nervosa. Mol Psychiatry 2013.
217 Appendix
44. Gendall K, Joyce P, Carter F, McIntosh V, Jordan J, Bulik C. The psychobiology and diagnostic significance of amenorrhea in patients with anorexia nervosa. Fertil Steril 2006; 85: 1531-1535. 45. Pinheiro A, Thornton L, Plotonicov K, Tozzi T, Klump K, Berrettini W, Brandt H, Crawford S, Crow S, Fichter M, Goldman D, Halmi K, Johnson C, Kaplan A, Keel P, LaVia M, Mitchell J, Rotondo A, Strober M, Treasure J, Woodside D, Kaye W, Bulik C. Patterns of menstrual disturbance in eating disorders. Int J Eat Disord 2007; 40: 424–434. 46. Tozzi F, Thornton L, Klump K, Bulik C, Fichter M, Halmi K, Kaplan A, Strober M, Woodside D, Crow S, Mitchell J, Rotondo A, Mauri M, Cassano C, Keel P, Plotnicov K, Pollice C, Lilenfeld L, Berrettini W, Kaye W. Symptom fluctuation in eating disorders: correlates of diagnostic crossover. Am J Psychiatry 2005; 162: 732-740. 47. Howie BN, Donnelly P, Marchini J. A flexible and accurate genotype imputation method for the next generation of genome-wide association studies. PLoS Genet 2009; 5: e1000529. 48. Altshuler DM, Gibbs RA, Peltonen L, Dermitzakis E, Schaffner SF, Yu F, Bonnen PE, de Bakker PI, Deloukas P, Gabriel SB, Gwilliam R, Hunt S, Inouye M, Jia X, Palotie A, Parkin M, Whittaker P, Chang K, Hawes A, Lewis LR, Ren Y, Wheeler D, Muzny DM, Barnes C, Darvishi K, Hurles M, Korn JM, Kristiansson K, Lee C, McCarrol SA et al. Integrating common and rare genetic variation in diverse human populations. Nature 2010; 467: 52-58. 49. Marchini J, Howie B. Genotype imputation for genome-wide association studies. Nat Rev Genet 2010; 11: 499-511. 50. Magi R, Morris AP. GWAMA: software for genome-wide association meta-analysis. BMC Bioinformatics 2010; 11: 288. 51. Morris JA, Randall JC, Maller JB, Barrett JC. Evoker: a visualization tool for genotype intensity data. Bioinformatics 2010; 26: 1786-1787. 52. Huckins L, Boraska V, Franklin C, Floyd J, Southam L, Nervosa GCfA, 3 WTCCC, Sullivan P, Bulik C, Collier D, Tyler-Smith C, Zeggini E, Tachmazidou I. Using ancestry-informative markers to identify fine
218 Appendix
structure across 15 populations of European origin Eur J Hum Genet in press. 53. Wang K, Zhang H, Bloss CS, Duvvuri V, Kaye W, Schork NJ, Berrettini W, Hakonarson H, Price Foundation Collaborative Group. A genome- wide association study on common SNPs and rare CNVs in anorexia nervosa. Mol Psychiatry 2011; 16: 949-959. 54. Aulchenko YS, Ripke S, Isaacs A, van Duijn CM. GenABEL: an R library for genome-wide association analysis. Bioinformatics 2007; 23: 1294- 1296. 55. Zhou X, Stephens M. Genome-wide efficient mixed-model analysis for association studies. Nat Genet 2012; 44: 821-824. 56. Gauderman WJ. Candidate gene association analysis for a quantitative trait, using parent-offspring trios. Genet Epidemiol 2003; 25: 327-338. 57. R: A language and environment for statistical computing. http://www.R-project.org, 2008. 58. Yang J, Lee SH, Goddard ME, Visscher PM. GCTA: a tool for genome- wide complex trait analysis. Am J Hum Genet 2011; 88: 76-82. 59. Boraska V, Davis OS, Cherkas LF, Helder SG, Harris J, Krug I, Liao TP, Treasure J, Ntalla I, Karhunen L, Keski-Rahkonen A, Christakopoulou D, Raevuori A, Shin SY, Dedoussis GV, Kaprio J, Soranzo N, Spector TD, Collier DA, Zeggini E. Genome-wide association analysis of eating disorder-related symptoms, behaviors, and personality traits. Am J Med Genet B Neuropsychiatr Genet 2012; 159B: 803-811. 60. Wade T, Gordon, S, Medland, Bulik, CM, Heath, A, Montgomery, GW, Martin, NG. Genetic variants associated with disordered eating. Int J Eat Disord 2013; 46: 594-608. 61. Fall T, Ingelsson E. Genome-wide association studies of obesity and metabolic syndrome. Mol Cell Endocrinol 2012. 62. Guo Y, Lanktree MB, Taylor KC, Hakonarson H, Lange LA, Keating BJ. Gene-centric meta-analyses of 108 912 individuals confirm known
219 Appendix
body mass index loci and reveal three novel signals. Hum Mol Genet 2013; 22: 184-201. 63. Neale BM, Medland SE, Ripke S, Asherson P, Franke B, Lesch KP, Faraone SV, Nguyen TT, Schafer H, Holmans P, Daly M, Steinhausen HC, Freitag C, Reif A, Renner TJ, Romanos M, Romanos J, Walitza S, Warnke A, Meyer J, Palmason H, Buitelaar J, Vasquez AA, Lambregts- Rommelse N, Gill M, Anney RJ, Langely K, O'Donovan M, Williams N, Owen M et al. Meta-analysis of genome-wide association studies of attention-deficit/hyperactivity disorder. J Am Acad Child Adolesc Psychiatry 2010; 49: 884-897. 64. Sklar P, Ripke S, Scott LJ, Andreassen OA, Cichon S, Craddock N, Edenberg HJ, Nurnberger JI, Jr., Rietschel M, Blackwood D, Corvin A, Flickinger M, Guan W, Mattingsdal M, McQuillin A, Kwan P, Wienker TF, Daly M, Dudbridge F, Holmans PA, Lin D, Burmeister M, Greenwood TA, Hamshere ML, Muglia P, Smith EN, Zandi PP, Nievergelt CM, McKinney R, Shilling PD et al. Large-scale genome- wide association analysis of bipolar disorder identifies a new susceptibility locus near ODZ4. Nat Genet 2011; 43: 977-983. 65. Genome-wide association study identifies five new schizophrenia loci. Nat Genet 2011; 43: 969-976. 66. A mega-analysis of genome-wide association studies for major depressive disorder. Mol Psychiatry 2013; 18: 497-511. 67. Watson HJ, Bulik CM. Update on the treatment of anorexia nervosa: review of clinical trials, practice guidelines and emerging interventions. Psychol Med 2012: 1-24. 68. Gauderman WJ. Sample size requirements for association studies of gene-gene interaction. Am J Epidemiol 2002; 155: 478-484. 69. Speliotes EK, Willer CJ, Berndt SI, Monda KL, Thorleifsson G, Jackson AU, Lango Allen H, Lindgren CM, Luan J, Magi R, Randall JC, Vedantam S, Winkler TW, Qi L, Workalemahu T, Heid IM, Steinthorsdottir V, Stringham HM, Weedon MN, Wheeler E, Wood AR, Ferreira T, Weyant RJ, Segre AV, Estrada K, Liang L, Nemesh J,
220 Appendix
Park JH, Gustafsson S, Kilpelainen TO et al. Association analyses of 249,796 individuals reveal 18 new loci associated with body mass index. Nat Genet 2010; 42: 937-948. 70. Hinney A, Hebebrand J. Three at one swoop! Obes Facts 2009; 2: 3-8. 71. Bradfield JP, Taal HR, Timpson NJ, Scherag A, Lecoeur C, Warrington NM, Hypponen E, Holst C, Valcarcel B, Thiering E, Salem RM, Schumacher FR, Cousminer DL, Sleiman PM, Zhao J, Berkowitz RI, Vimaleswaran KS, Jarick I, Pennell CE, Evans DM, St Pourcain B, Berry DJ, Mook-Kanamori DO, Hofman A, Rivadeneira F, Uitterlinden AG, van Duijn CM, van der Valk RJ, de Jongste JC, Postma DS et al. A genome-wide association meta-analysis identifies new childhood obesity loci. Nat Genet 2012; 44: 526-531. 72. Scherag A, Dina C, Hinney A, Vatin V, Scherag S, Vogel CI, Muller TD, Grallert H, Wichmann HE, Balkau B, Heude B, Jarvelin MR, Hartikainen AL, Levy-Marchal C, Weill J, Delplanque J, Korner A, Kiess W, Kovacs P, Rayner NW, Prokopenko I, McCarthy MI, Schafer H, Jarick I, Boeing H, Fisher E, Reinehr T, Heinrich J, Rzehak P, Berdel D et al. Two new loci for body-weight regulation identified in a joint analysis of genome-wide association studies for early-onset extreme obesity in French and German study groups. PLoS Genet 2010; 6: e1000916. 73. Visscher PM, Brown MA, McCarthy MI, Yang J. Five years of GWAS discovery. Am J Hum Genet 2012; 90: 7-24. 74. Sullivan PF, Daly MJ, O'Donovan M. Genetic architectures of psychiatric disorders: the emerging picture and its implications. Nat Rev Genet 2012; 13: 537-551. 75. Ripke S, O'Dushlaine C, Chambert K, Moran J, Kähler A, Akterin S, Bergen S, Collins A, Crowley J, Fromer M, Kim Y, Lee S, Magnusson P, Sanchez N, Stahl E, Williams S, Wray N, Xia K, Bettella F, Børglum A, Cormican P, Craddock N, de Leeuw C, Durmishi N, Gill M, Golimbet V, Hamshere ML, Holmans P, Hougaard D, Kendler K et al. Genome-
221 Appendix
wide association analysis identifies 13 new risk loci for schizophrenia. Nat Genet 2013: 45:1150-1159.
222 Appendix
Table 1. List of ethnicities and numbers of samples for main case control and anorexia nervosa (AN) subtype analyses across discovery and replication datasets
Cases AN AN BINGE- Country (% of RESTRICTING PURGE subtype Controls (% of females) subtype cases cases females) Discovery dataset*
Canada 54 24 25 417 (46.52) Czech Republic 72 40 29 331 (35.04) Finland 131 39 29 404 (100) France 293 137 135 619 (60.09) Germany 475 147 55 1,205 (49.13) Greece 70 10 5 79 (100) Italy-North 203 103 99 841 (52.19) Italy-South 75 31 26 52 (100) Netherlands 348 115 90 593 (51.26) Norway 82 24 15 602 (67.44) Poland 175 68 107 564 (29.43) Spain 186 45 44 185 (75.14) Sweden 39 28 11 975 (72.10) UK 213 97 97 5,163 (49.43) USA 491 311 165 2,830 (41.31) Total discovery 2,907 1,219 932 14,860 (51.73) In silico replication
USA-Hakonarson 1,033 (97.67) 0 0 3,775 (45.85) Estonia 31 (100) 0 0 106 (100) De novo replication
Austria 48 (100) 0 0 183 (65.03) Czech Republic 32 (71.88) 0 0 22 (100) Finland 15 (100) 0 0 94 (8.51) France 55 (100) 0 0 123 (100) Germany 174 (99.43) 31 64 380 (66.84) Greece 16 (100) 0 0 53 (100) Italy-South 156 (96.79) 32 24 63 (100) Netherlands 229 (100) 45 23 380 (27.11)
223 Appendix
Poland 52 (98.08) 0 0 93 (100) Spain 10 (100) 0 0 328 (41.46) UK 155 (100) 28 55 199 (65.83) USA** 671 (100) 349 272 2,830 (41.31) Japan 458 (100) 213 240 421 (100) Total replication 3,135 (98.72) 698 678 9,050 (50.08) Total global meta- 5,551 1,606 1,445 21,080 analysis *All AN cases from discovery dataset were females. **USA samples from discovery dataset were merged together with USA replication samples for replication analysis. The same USA control dataset was used.
Table 2: Global meta-analysis results of SNPs with the greatest evidence of association for the main anorexia nervosa (AN) case-control analysis
SNP information Global meta-analysis across discovery and replication datasets 2 CHR POS MARKER NEAREST GENE EA NEA EAF OR OR_95LOR_95U P I N_st N_sa 3 182794261 rs9839776 SOX2OT T C 0.270 1.158 1.095 1.225 3.01E-07 0 27 21857 4 102267099 rs17030795 PPP3CA G A 0.192 1.149 1.082 1.220 5.84E-06 0 24 23111 8 19584542 rs11204064 CSGALNACT1 G A 0.477 1.118 1.063 1.176 1.57E-05 0.008 28 21477 3 12013264 rs2618405 7.5kb from SYN2 C A 0.218 1.152 1.079 1.229 2.03E-05 0.244 22 18566 13 23433988 rs1886797 18kb from SPATA13 T C 0.301 1.133 1.070 1.200 2.18E-05 0.317 25 15827 21 21257379 rs10482915 35kb from NCAM2 A G 0.074 1.193 1.097 1.297 3.96E-05 0 28 26164 7 106473684 rs2395833 PRKAR2B T G 0.334 1.101 1.051 1.154 5.62E-05 0.132 29 26511 2 80768625 rs1370339 39kb from CTNNA2 C T 0.472 1.098 1.049 1.149 5.68E-05 0 29 26508 13 63470128 rs9539891 255kb from OR7E156P C T 0.332 0.891 0.842 0.942 5.88E-05 0 23 20389 2 225017222 rs1523921 26kb from CUL3 / 42kb from FAM124B T C 0.210 1.131 1.065 1.201 5.95E-05 0.162 26 21858 19 11650015 rs206863 ZNF833P A G 0.899 0.864 0.804 0.928 6.47E-05 0.076 28 26402 23 107578961 rs5929098 COL4A5 T C 0.771 1.135 1.066 1.210 8.37E-05 0.002 29 19249 7 146565029 rs6943628 CNTNAP2 A G 0.097 1.161 1.077 1.251 9.38E-05 0 29 26377 CHR - chromosome; POS - position in hg18; EA - effect allele; NEA - non- effect allele; EAF - effect allele frequency; OR - odds ratio; OR_95L - lower 95% confidence interval; OR_95U - upper 95% confidence interval; P - p- value; I2 - measure of heterogeneity; N_st - number of contributing studies; N_sa - number of contributing samples.
224 Appendix
Figure 1: Analysis of RNAseq data for whole-brain tissue obtained from 12 different mouse strains for 32 mouse orthologues of the 34 human genes for which association to anorexia nervosa (AN) was identified. The average FPKM (Fragments Per Kilobase of exon per Million fragments mapped) values for 32 genes across 12 mouse strains are shown.
225 Chapter 8
Chapter 8. Discussion and conclusions
Overview of genetic research in anorexia nervosa: the past, the present and the future
Marek K. Brandys Carolien G.F. de Kovel Martien J. Kas Annemarie A. van Elburg Roger A.H. Adan
International Journal of Eating Disorders 2015 Nov; 48(7):814-825
226 Chapter 8
Abstract
Background: Even though the evidence supporting the presence of a heritable component in the aetiology of anorexia nervosa (AN) is strong, the underlying genetic mechanisms remain poorly understood. The recent publication of a genome-wide association study (GWAS) of AN 1 was an important step in genetic research in AN. Objective: To briefly sum up strengths and weaknesses of candidate-gene and genome-wide approaches, to discuss the genome-wide association studies of AN and to make predictions about the genetic architecture of AN by comparing it to that of schizophrenia (since the diseases share some similarities and genetic research in schizophrenia is more advanced). Method: Descriptive literature review. Results: Despite remarkable efforts, the gene-association studies in AN did not advance our knowledge as much as had been hoped, although some results still await replication. Discussion: Continuous effort of participants, clinicians and researchers remains necessary to ensure that genetic research in AN follows a similarly successful path as in schizophrenia. Identification of genetic susceptibility loci provides a basis for follow-up studies.
227 Chapter 8
Unravelling of the background of polygenic psychiatric disorders, such as AN, turned out to be more challenging than it had been envisioned in the early days of the genetic research in psychiatry. Nevertheless, recent advancements in schizophrenia's (SCZ) research set an encouraging example. The present paper overviews the past approaches to unravel the genetics underlying AN and discusses the current knowledge about the genetic architecture of AN by relating it to a field of psychiatry with more advanced research in genetics, i.e. SCZ. This discussion is particularly timely, as the largest genome-wide association study (GWAS) of AN to date has been published recently 1, and the need for evaluation of past, present and future approaches is evident.
Heritability and rationale for gene-association studies
Several lines of evidence suggest that there is a substantial genetic component in the aetiology of AN. AN has been observed across many cultures 2. Strong familiar aggregation of AN has been documented (relative risk of 11.3 in first-degree relatives of cases with AN, as compared to the general population 3,4), and the heritability (h2) has been estimated in several twin studies and one adoption study of disordered eating symptoms 5. These estimates range from 0.56 (95% CI, 0.00-0.87) 6 to 0.74 (95% CI: 0.35-0.95) 7, depending on the studied population, definition of AN and applied methodology. Thus, a genetic component in the liability to AN has been demonstrated, although interpretation of h2 can be problematic. h2 is an estimate of a fraction of phenotypic variance that can be attributed to the genetic variance. These estimates are often cited in papers as a rationale for embarking on a gene-association study (in a way that a high h2 is supposed to suggest larger contribution of genes to the aetiology of the disease 8). However, it needs to be remembered that h2 is a relative value. It remains valid for the population under study, at a given time; when environmental factors are altered, h2 changes as well. For instance, in a
228 Chapter 8 population at risk of developing AN, such as ballet dancers 9, where the variation in exposure to environmental risk factors is low (so the environment is more uniform), the estimate of AN’s h2 will be higher than in the general population (high h2 does not mean that the environmental factors are less important in the aetiology). h2 informs about the proportion of the variation in a population that is due to genetic factors, and not about the fraction of cases attributable to genetic factors 10. The exact value of h2 is not as important as its qualitative interpretation. The h2 estimate provides a hint about whether there is detectable genetic variance in a given trait 8, but it does not say anything about how many genes might be involved, what their impact might be or whether it will be easy or not to identify the underlying variants. With the advent of genome-wide genotype data it has become possible to estimate how much of the genetic variance is explained solely by SNPs captured on genotyping microarrays 11. This so-called SNP-heritability is more informative in terms of the genetic architecture of a disease/trait than the h2 calculated from the family studies. Unfortunately, to date it has not been possible to generate such an estimate for AN. Nevertheless, aggregated evidence coming from several lines of research hints that genetic factors are pivotal in the aetiology of AN. No monogenic forms of AN have been found and the data suggest that the genetic underpinning of AN is multifactorial (i.e. multiple genetic variants with small effects, rather than one or a few potent variants, working in concert with environmental factors) 12. Two types of studies have been employed in search for those genetic factors. The linkage approach, which investigates co-segregation of genetic regions with the disease status in large families, has been successful in detecting rare and very potent genetic variants involved in aetiology of single-gene disorders (Mendelian), e.g. cystic fibrosis or Huntington’s disease 13,14. However, its usefulness in unravelling common variants of small effects in complex, polygenic diseases or traits remains very limited.
229 Chapter 8
The second category is a population-based genetic-association study, which investigates whether frequencies of certain genotypes or alleles are different between cases and controls (significant difference implies association) or if they are correlated with a quantitative trait. This approach focuses on variants with small or medium effects, in a multifactorial model. Within this category, candidate-gene studies (CGSs) look into single- nucleotide polymorphisms (SNPs) in biologically plausible genes, whereas GWAS test common SNPs distributed throughout the whole genome. Although this category of studies uses data of unrelated individuals most often, inclusion of family trios or siblings is possible, as long as the level of relatedness is known to the researcher.
Candidate gene approach
The candidate-gene approach in AN, much like in other psychiatric disorders, turned out to be a primarily futile effort. The scarcity of successful replications can be explained by several reasons, such as genetic differences between the discovery population and the populations in the replication attempts, or by errors and biases leading to false positive results. These potential errors include: • Imperfect matching of cases to controls in terms of the ethnic background. Different genetic backgrounds of cases and controls (population stratification) might lead to spurious associations. The CGSs offer no ways of attenuating this risk, except for relying on declarations of participants or inclusion of ancestry-informative markers 15 (the latter, however, is a relatively recent method, and it was not commonly applied when the most of the CGSs were performed). • Some degree of relatedness between participants may lead to spurious associations, if it is not taken into account in the analysis.
230 Chapter 8
• Technical biases might occur if DNA samples of cases and controls come from different sources (blood vs buccal swabs). • Various techniques of genotyping are prone to error, and ideally the samples of cases and controls should be randomly distributed across plates and genotyped under the same conditions. The winner’s curse might also be partially responsible for the lack of replications. It is a statistical phenomenon of regression to the mean – the study which is first to identify a genetic association might find a larger effect size than it is in reality. Thus, the subsequent studies might not confirm this association, because even if it is true, its effect is likely to be smaller than in the first published study (this is also often observed in multi-stage GWASs), and therefore the samples that are used in the replication studies are too small to detect the effect. Overly optimistic expectations towards the potential effects of associated alleles might have led to overestimation of statistical power (a probability of not missing a true association), and the publication pressure might have resulted in the lack of appreciation for proper adjustment of results for multiple comparisons 16. Finally, although, the CGSs focused on genes, often the common variation within the gene or adjacent to it was not well captured, due to insufficient SNP coverage (i.e. a density of probes throughout a genetic loci was not sufficient to cover all of the common SNPs). Retrospectively, given the complexity and redundancy of biological pathways, and in light of what is now known about the genetic architecture of psychiatric diseases, the hypotheses about which genes could potentially harbor causative mutations had small chances to be proven right. Out of hundreds of associations indicated by CGSs in biomedical research only a few were replicated in GWASs 17. This ratio is even less favourable in the field of psychiatry. A study by 18 found lack of enrichment of association signal in a large genome-wide dataset of cases with schizophrenia and controls after analysis of 732 autosomal genes indicated in 1374 CGSs (investigation of signal enrichment involves collective testing of a selected group of variants in
231 Chapter 8 an independent dataset; it has much greater power, compared to testing of individual variants).
Candidate gene studies in anorexia nervosa
Comprehensive reviews of CGSs in AN are available elsewhere 19,20. Although the selection of candidate genes for studies of AN was based on interesting hypotheses 21, and more than 200 gene-association studies were performed in the context of EDs, up to date none of the initially promising findings have been convincingly replicated in the subsequent candidate or genome-wide studies. Meta-analyses, which summarized and weighted the evidence from multiple studies, were also disillusioning 22-25. Also the relatively recent CGS which used the modern standards of design, quality control and statistical significance was negative 26. Still, there are a few findings which await replication attempts, such as rs1473473 of TPH2 27, the 5-HTTLPR polymorphism on SLC6A4 28, rs7180942 in NTRK3 29 and Ala67Thr variant in AGRP 30 (these polymorphisms were not tested in two recent GWASs of AN, because they were not present on the genotyping arrays used in those studies). In parallel to the growing disillusionment about the candidate-gene method, a new approach towards investigation of genetic associations emerged. GWAS technology is relatively recent (first GWAS dates back to 2005 31), but it already has had significant impact on the landscape of biomedical research and resulted in progression of aetiological knowledge about diseases and traits 32.
Genome-wide association approach
GWAS is a hypothesis-free approach. It uses microarray platforms to examine the genotypic data from a large number of SNPs (from hundreds of thousands up to millions), which cover most of the human common SNP
232 Chapter 8 variation (a SNP is considered common if the frequency of its minor allele is larger than 1%). This coverage is increased via imputation - a procedure which uses statistical algorithms to infer the genotypes of the ungenotyped SNPs by employing the reference data coming from e.g. HapMap or 1000 Genomes Project populations. Genome-wide data also allows for investigation of copy number variants (CNVs; deleted or duplicated stretches of the genome).
Below is a list of the main goals of GWASs: • Furthering the understanding of the biological mechanisms of the disease, by finding the genes and pathways involved in the aetiology. This is the foremost goal of GWASs. • Learning about the genetic architecture. This includes the expected range of effect sizes, allelic frequencies of the associated variants, underlying genetic models (additive, dominant, recessive, overdominant, multiplicative) and the possibility of gene x environment and gene x gene interactions. • Understanding of the genetic overlap between diseases and traits. This has a potential of enhancing the nosological system and treatment. • Genetic screening to identify populations at risk (risk prediction) or individual genotyping of a patient to inform diagnosis and treatment (personalized medicine). As exciting as these prospects are, they are distant goals, and in view of a highly polygenic nature of psychiatric diseases, they are unlikely to be achievable in the near future 33.
There are many potential sources of systematic errors in these studies, much like in CGSs, but the nature of genome-wide data provides opportunity to detect such errors, and, in some cases, adjust for them. In particular, the danger of obtaining false results due to a mismatch between cases and controls (e.g. due to population stratification or a batch bias) can be identified and partially controlled for. Additionally, various quality control
233 Chapter 8 procedures may aid in the elimination of technical artifacts or covertly related participants. Reliability of results can be verified by post-hoc analyses, such as the LD-score regression approach 34. Furthermore, it has become a standard for modern GWASs to include replication of top signals in independent samples. Large numbers of statistical tests being carried out in GWASs mean that many of them will reject the null hypothesis due to chance. For example, for 500k independent tests for association, 25k tests will achieve significance of p<0.05, even if they are not associated (leading to a type-I, false positive error). Alpha needs to be adjusted for multiple comparisons - a level of significance accepted as reliable in GWASs with approximately 1 million independent tests is p<5x10-8. As a consequence, a true signal of association might be obscured in the midst of statistical noise (a type-II error - false- negative), especially since the effects of true associations in the studies of complex phenotypes tend to be small. Odds ratios (OR) are commonly used as a measure of effect size in GWASs of binary phenotypes. For example, ORs of SNPs associated with schizophrenia (SCZ) are typically in a range between 1.05 and 1.2 (reciprocally –0.83 to 0.95), with single SNPs explaining from 0.05% (SNP rs1344706, nearest gene Znf804A) to 0.67% (SNP rs7341475, nearest gene RELN) variance in the disease (the amount of explained variance depends on SNPs effect size and allelic frequency in population) 35,36. The way to avoid the risk of false-negative errors in GWASs is increasing the sample size. The statistical power of a test informs about the likelihood of the type-II error (i.e. missing a true association). It is a function of the sample size, expected effect size and frequency of the tested allele in the sample. It also depends on assumptions about the genetic model underlying the association, and on the fact whether the tested variant is assumed to be causal or rather correlated with the causal variant. The power estimations in the earlier studies were often overly optimistic, as they were based on the hope of finding moderate-to-large effects (i.e. OR>1.25). If SNPs of common frequency and with moderate-to-large effects
234 Chapter 8 were to exist in AN, they would have likely been detected by now. Nowadays, it is known that the majority or all of the associated variants with common allelic frequencies will have small effect sizes. With the benefit of hindsight, it is clear that carrying-out an underpowered study or not including an independent replication is likely to lead to unreliable results, and, in fact, power as high as 90% is recommended 32. This means that in the field of psychiatry tens of thousands of cases are necessary for a GWAS to succeed (Fig. 1). What needs to be remembered when interpreting a GWAS is that its results inform about association but do not determine causality, and that a statistical strength of association at a given locus should not be confused with its biological relevance (the most significant finding in GWASs might not be the most informative).
Fig. 1. Number of cases required for 80% power in relation to allele frequency, for different sizes of genetic effect, expressed in odds ratios, assuming an additive model of genetic effect. The calculation assumes 3:1 control:case ratio, 1% frequency of the disease in the general population and a genome-wide significance level (α=5*e-8). Calculated with Quanto 1.2.4 91.
235 Chapter 8
Genome-wide association studies in anorexia nervosa
The first genome-wide study of AN was published in 2009 37. It was based on a DNA-pooling approach, in which allele frequencies were estimated from pools of DNA of cases vs controls (in contrast to classical individual genotyping in cases and controls). The authors used a set of microsatellite markers to search for genetic loci associated with AN in groups of Japanese cases and controls. This method has several disadvantages 38. Most importantly, it does not provide a way to guard against population stratification and does not allow inclusion of covariates. The analysis of microsatellite markers led to identification of 10 potentially associated loci. The second stage of the study was a single SNP fine-mapping analysis of indicated regions (96.6% of 331 cases with AN overlapped with the first stage of the study), which found a SNP rs2048332 (downstream of SPATA17 gene) to be most strongly associated with AN (p-value=0.0001). However, these interesting results need to be viewed in the context of several limitations, such as using a technique which is prone to errors (DNA-pooling), small sample sizes in both stages of the study and lack of control for population stratification and cryptic relatedness (i.e. relatedness between participants unknown to the researcher and not taken into account in the analysis). The implication of SPATA17 was not replicated by two later GWA studies in AN, but the most significant SNP from the study of Nakabayashi et al. was not present on their genotyping microarrays (also, those studies were performed on participants with European descent and they used different methodology).
In 2011 Wang et al. published a GWAS of a sample of 1033 cases with AN (98% female) and 3733 pediatric controls (46% females) 39. In retrospect, the fact that this study found no genome-wide significant associations comes as no surprise. Underpowered as for the current standards, it was unable to detect associations with small effect sizes. With this sample size, the study had 80% power to detect an association of a SNP
236 Chapter 8 with minor allele frequency of 10% and OR>1.58, at genome-wide significance, whereas the ORs typically observed in the GWASs of psychiatric disorders (with a few exceptions) are below 1.25 32. The control group was composed of pediatric subjects, which could have decreased the power even further, since some of the children might develop AN or other psychiatric disorder later in life. Additionally, the genome-wide genotype data allowed the authors to test a hypothesis whether rare and large CNVs, which were previously associated with several psychiatric disorders 40, associate with AN. None of those CNVs were found to associate with AN, but again, insufficient sample size was a major limitation. Even though some of those CNVs were found to have ORs ranging from 5-20 in SCZ 32, their extremely low frequencies drastically decrease statistical power.
Recently, the Genetic Consortium for Anorexia Nervosa (GCAN) and the Wellcome Trust Case Control Consortium 3 performed a GWAS of AN which included 2907 female cases with AN and 14860 mixed-sex in silico controls in the first stage of the study, and 2677 cases and 8629 controls in the replication phase 1. In silico, in this context, means that the controls were not genotyped along with the cases but retrieved from the already existing databases, which reduced the overall costs of genotyping, but came at the price of difficulties in the analysis and interpretation. Cases and controls came from 14 countries and they were ancestrally matched to each other within multiple strata, to protect against population stratification. After the discovery stage of the study, 76 most promising SNPs were carried forward (prioritized) to a replication stage in an independent sample. None of those SNPs reached genome-wide significance. The two most strongly associated variants were rs9839776 in SOX2OT (p=3.01x10-7) and rs17030795 in PPP3CA (p=5.84x10-6). Subtype analyses were also carried out, since AN is categorized into AN-restricting type and AN-binging/purging type (both types are underweight and restrict calorie-intake but the former does not have binging
237 Chapter 8 episodes and do not engage in purging, such as vomiting or use of laxatives). The hope behind these analyses was that the benefit of increased phenotypic homogeneity of the subsets would outweigh the loss of power due to decreased sample size. Subtyping worked well in, for instance, a GWAS of ischaemic stroke 41. In this case, however, the analyses did not reveal any variants with association signals stronger than in the main analysis. Subtype analyses might offer a promising approach, but not before the size of the sample becomes sufficient to detect small effect sizes. Several important conclusions came from this study. A sign test was performed to determine whether alleles of SNPs present in both the discovery and the replication stage showed the same direction of effect in both stages (i.e. if they increase or decrease the risk for a given phenotype). In case of no enrichment of association signal among the prioritized SNPs, 50% of considered SNPs should have the same direction of effect in both stages. It was demonstrated via a sign test that there was a true signal of association among the 76 prioritized SNPs, but the statistical power of the study was insufficient to detect it on a level of individual SNPs. This means that the effect sizes of truly associated SNPs are expected to be small. Nevertheless, the term “suggestive association of a SNP” in an underpowered GWAS should be taken with great caution. Not only does low power reduce the chances of discovering a true effect but it also decreases the likelihood that a nominally significant finding reflects a true association 42. Nine variants with the lowest p-values found in Wang et al. (2011) and 12 variants from the studies of ED-related traits 43 were tested in this GWAS, but they showed no evidence of association (individually, or collectively via a sign test). The findings reported in those earlier studies were all below genome- wide significance. Another observation concerned the fact that 60 SNPs associated with four psychiatric disorders (SCZ, attention deficit hyperactivity disorder (ADHD), bipolar disorder (BD) and major depressive disorder (MDD)) showed no enrichment of the association signal in the genome-wide data of cases with AN and controls 1. Similarly, there was no enrichment of the signal for
238 Chapter 8
89 SNPs known to associate with BMI in the general population 44,45. Interestingly, evidence of slight enrichment was found for 13 SNPs previously associated with morbid obesity 44, which suggests some overlap in the genetic aetiology of AN and the opposite, extreme end of the BMI spectrum (but not the non-extreme part of this spectrum). These analyses provide no evidence for genetic overlap between determinants of AN and other psychiatric disorders or AN and BMI (in a normal range), but they were not sufficient to claim that such overlap does not exist at the level of common variation. More comprehensive ways of testing for genetic overlap will become possible when much larger sets of genome-wide data of individuals with AN become available. To sum up, the outcomes of this largest up-to-date GWAS in AN were only modestly informative. However, the success of a genome-wide approach in a given disease cannot be judged fairly as long as the sample size is not sufficiently large 46.
Recent progress of gene-association studies in schizophrenia
GWAS technology is relatively recent (the first GWAS dates back to 2005 31), but it already has had a tremendous impact on the landscape of biomedical research 32. For instance, the identification of inflammatory pathway regulation in macular degeneration 31 or autophagy pathway in inflammatory bowel disease 47 are examples of how GWASs implicated unexpected biological mechanisms in the aetiology of diseases. In psychiatric disorders, the genetics of SCZ is arguably the most advanced field. By referring to this encouraging example, we intend to discuss what the future of the genetic studies in AN might be. The two recent GWASs of SCZ have led to remarkable progress in the understanding of the genetics of SCZ 48,49. A study from 2013, which combined the data of 21k cases and 38k controls, reported twenty-two
239 Chapter 8 robust association signals. These findings gave basis to follow-up analyses, where a gene or a pathway became a unit of analysis, rather than individual SNPs. Although a gene-association study does not determine causality, it can indicate it, especially when the number of independently associated SNPs becomes larger. Four main implications about the aetiology of SCZ came from this study and they pertain to calcium channel genes (also implied in autistic spectrum disorder - ASD, and BD 48), to the major histocompatibility complex region of the genome (MHC, fundamental for the immune system), to the mRNA-137 pathway and to the long intergenic non-coding RNAs. It is a good example of how a successful GWAS advances biological knowledge and generates targets for subsequent studies. This was possible even though the 22 significant association signals constitute only a tiny fraction of the population of independent SNPs expected to be associated with SCZ. Using these genome-wide data, Ripke et al. (2013) performed analyses which delineated the genetic architecture of SCZ. By analysing the genetic similarities between case-case, case-control and control-control pairs (testing whether cases are genetically more similar to each other than to controls 50) the authors were able to estimate the SNP-heritability of SCZ (i.e. a portion of heritability explained solely by SNPs) to be 32%. They also report that about 80% of all SNPs associated with SCZ are expected to have a frequency larger than 1% 11. An alternative method 51 produced a slightly higher estimate of SNP-heritability and indicated that 8300 SNPs (6300- 10200 95% CI) contribute to the genetic underpinnings of SCZ. This method also projected that for 60k cases and 60k controls about 794 independent SNP associations would be expected. These analyses should be interpreted with caution, as they rely on various assumptions – their exact numerical output is not as important as the conclusion that a substantial proportion of the genetic heritability to SCZ is explained by thousands of SNPs with small effect sizes and predominantly in the range of common frequency. The most recent GWAS of SCZ corresponds with the findings of the study discussed above 49. In an unprecedented effort to analyse the data of almost 37k cases and over 113k controls, the authors confirm that increasing
240 Chapter 8 the sample size leads to new findings (128 independent associations, ascribed to 108 genetic loci, 83 of which reported for the first time in this study). Some of the associations point towards genes and regions previously implicated in GWAS and rare-variant studies (voltage-gated calcium channel subunits, MHC, genes involved in glutamatergic neurotransmission and synaptic plasticity 48). Other findings suggest aetiological involvement of G- protein coupled receptor genes (including DRD2 gene, a target of antipsychotic medications) and genes related to other ion channels and to neurodevelopment. To some extent, the implications of these results converge with the proposed aetiological hypotheses of SCZ (although, it should be stressed again that GWAS do not determine causality). This is still the beginning of a long way to reveal a substantial portion of SNPs associated with SCZ. Nevertheless, these encouraging results have already led to meaningful biological inferences and important follow-up studies. For instance, it has been shown that expression of the ZNF804A gene in the dorsolateral prefrontal cortex is dependent on the genotype of SNP rs1344706 (previously associated with SCZ) 52. Also, the availability of genome-wide data for several psychiatric diseases, including SCZ, made it possible to study their genetic overlap, which led to discovery that the genetic diathesis of SCZ have many points of convergence with ASD, BD and other psychiatric disorders 53. Interestingly enough, a recent study found positive genetic correlation between SCZ and AN 54.
Genetic architecture of anorexia nervosa, as compared to schizophrenia
The genetic landscape of AN has been explored less than that of SCZ. The authors of the recent GWAS of AN attempted to estimate the SNP- heritability (similar to what Ripke et al. did in the SCZ study 48), but the results of this analysis were not judged reliable - the in silico controls were matched to the cases within multiple small-sized strata, which did not
241 Chapter 8 provide a solid basis for this estimation 1. Therefore, we will discuss the genetic architecture of AN based on indirect premises, such as its epidemiological characteristics, and comparisons with SCZ. Estimates of lifetime prevalence of AN in women range from 0.9% to 2.2% 55,56, whereas the lifetime prevalence of SCZ (for both sexes) is estimated to be between 0.4% 57 and 1.6%58,59. Both diseases have high mortality rates. The standardized mortality ratio (SMR) is very high in AN (SMR of 6, i.e. nearly 6 times greater mortality than in the general population 60), whereas SMR in SCZ is 2.6 61). AN and SCZ have generally an early age of onset (median ages of onset are 15 and 22, respectively 62), but the pubertal period appears to be pivotal in AN 63. The affected individuals have markedly reduced fertility ratio (FR; measured as a number of children in comparison to the general population); this reduction is greater in SCZ (FRmen=0.23 and 59 FRwomen=0.47) than in AN (FRmen=0.54 and FRwomen=0.81) . Thus, both diseases are associated with survival and reproductive disadvantage. Also, similarly to SCZ, no Mendelian forms of AN have been identified thus far (unlike ASD, Alzheimer’s disease and mental retardation). Altogether, it suggests that common variants with moderate or strong effects are unlikely to be involved in the aetiology of those disorders, because such variants are pruned out of the genetic pool due to the evolutionary pressures (purifying selection). On the other hand, common and rare variants with small effects remain nearly invisible to purifying selection (they "behave" like neutral mutations), especially in the light of the recent population expansion 64. Their individual effects are almost negligible, but collectively such variants are responsible for a substantial portion of the risk to the disease. This model fits well with what has been observed in the GWASs of SCZ and, with less certainty due to scarcity of data, in AN. An alternative model posits that the susceptibility variants remain in the population because they have ambivalent effects on fitness (e.g. they used to increase fitness under specific environmental circumstances in the evolutionary past but are deleterious in the modern ages or they are deleterious in the affected individuals but might have beneficial effects in their unaffected relatives). This model, which relies
242 Chapter 8 on balancing selection, has found less empirical support (siblings of individuals affected with SCZ or AN do not seem to have increased fecundity) 59,62), but it cannot be ruled out. It was demonstrated that de novo variants also contribute to the susceptibility to SCZ. This is known from the recent genetic studies using sequencing 65 and was also suspected on basis of the increased age of fathers of individuals with SCZ (in comparison to the general population) 66. The association between advanced paternal age and SCZ (and some other psychiatric disorders) is likely due to a higher rate of de novo mutations coming from the male germline 67. Whether this could also be true for AN remains an open question, due to the scarcity of data. One study found a positive correlation between paternal age and disordered eating in offspring, in particular in the case of fathers older than 40 years of age 68, which suggests a possible role of de novo mutations in the aetiology of AN. However, large-scale sequencing studies are required to verify that. SCZ has been robustly associated with several rare, large and recurrent CNVs, which are also known to associate with other psychiatric and non-psychiatric disorders. A possible role of those CNVs in AN remains unknown. One study attempted to test some of those CNVs in AN, and found no association, but in view of the insufficient power its conclusions were limited 39. Another observation concerns the fact that AN is far more frequently observed in females than in males (studies report the female:male ratio of lifetime prevalence to range between 3:1 55 to more than 10:1 69). This sexual dimorphism is stronger than in other psychiatric disorders and it might have consequences for the genetic architecture (e.g. a greater involvement of chromosome X or sex steroid-responsive genes in the aetiology). Eating disturbance and severe emaciation are distinguishing features of AN. The hope that due to apparent involvement of food-regulatory circuitry and measurable phenotype (body weight) the complexities of genetic underpinnings of AN would be revealed easier (i.e. the effect sizes would be larger and phenotypic heterogeneity lower), than in the case of
243 Chapter 8 other psychiatric disorders, did not prove right. All in all, it is reasonable to suspect that on a general level, the genetic architecture of AN is not fundamentally different from that of SCZ. Thousands of independently associated SNPs with small effect sizes and from the common range of allelic frequencies, which point towards hundreds of loci, are likely to account for a sizeable part of the variance in liability to AN. It is clear by now that there are no common variants of moderate-to-large effects (should they exist, they would have been found by now - the study by Boraska et al. had 80% power to detect association with OR=1.32 for a SNP with a minor allele frequency of 10%, at a genome-wide significant level). Associated variants are expected to be present across the whole range of allelic frequencies (de novo, rare and common). It remains unclear whether variants with very low allele frequencies might have larger effect sizes than the common variants 35.
Additional phenotypes and genetic overlap between disorders
Various degrees of genetic overlap were demonstrated across psychiatric entities, at the levels of de novo, rare and common variation. For instance, Fromer et al. 65 showed in a sequencing study of cases with SCZ and their parents that loss-of-function de novo mutations were enriched in cases with SCZ within the group of genes implicated into aetiology of autism. Rare variants, such as CNVs, are also known to associate with several psychiatric disorders and epilepsy 40. Purcell at al. showed that the aggregate polygenic contribution of many common alleles with small effects on the liability to SCZ (polygenic risk score) also increased the risk for BD 70. In another study, the genome-wide data of common variants allowed for investigation and indication of shared genetic aetiology in SCZ, BD, ASD, MDD and ADHD 71 (interestingly, calcium-channel activity genes emerged again and appeared to have pleiotropic effects on psychopathology) or in SCZ and multiple sclerosis 72. Other study quantified the extent of the overlap in genetic
244 Chapter 8 variation between disorders, and showed that it was substantial between e.g. SCZ and BD or SCZ and MDD 73. The analyses of genetic overlap between AN, other eating disorders and psychiatric disorders are likely to be performed in the future, but, for the moment, the number of genome-wide genotyped individuals with AN remains too small. These observations support the notion that psychiatric diagnoses require revision, since they do not reflect the recent advancement of our understanding of relationships between disorders. This problem is apparent in the field of EDs 74. In a longitudinal study, most of the patients with AN experienced diagnostic cross-over (more than half migrated between AN subtypes and 34% migrated to BN) 75. Also, eating disorder not-otherwise- specified is the most often established diagnosis in the clinical practice, meaning that the criteria for core diagnoses (AN, bulimia nervosa, binge- eating disorder) are failing at discriminating the majority of patients 76. Psychiatric comorbidities in EDs are ubiquitous. Depression and anxiety disorders are similarly frequent across ED categories, whereas diagnoses of OCD are more frequent in AN 77. ASD and ADHD are also observed among patients with AN 76. The specific diagnoses within the category of EDs are likely to have highly overlapping aetiologies, and aetiological mechanisms shared with other psychiatric disorders are also plausible 3. There exist cross-diagnostic phenotypes that reflect the general psychopathological domains of psychiatric diseases 65,78. Furthermore, it has already been suggested many years ago that psychiatric research and taxonomy could benefit from focusing on phenotypes which are intermediate between the genetic causes and the diagnoses (so-called endophenotypes or intermediate phenotypes (IP) 79. IP are thought to be quantitative traits, which are less complex, more accurately measured and closer to the genetic substrate than the diagnoses. In view of the substantial genetic overlap between several psychiatric disorders, it is plausible that the postulated domains of psychopathology represent such IP.
245 Chapter 8
Additional phenotyping in genetic studies of psychiatric disorders can be very useful in the future research, but it has drawbacks. Positive aspects of application of valid IP in genetic research include: • More reliable ascertainment of cases, due to higher accuracy of measurement (lesser phenotypic heterogeneity) • Possibility of ranking of individuals within diagnoses according to the IP level • Possibility of merging of groups of cases with different diagnoses, according to the IP level • Possibility of inclusion of covariates in the model • Possibility of inclusion of individuals from the general (non-clinical) population, according to the IP level • In general, tests of quantitative traits have higher power than tests of binary traits • Possibly, larger effect sizes of associated genetic variants • Simpler interpretation of results in terms of mechanisms of action.
On the other hand, the negative aspects include: • Practical difficulties with collecting large samples of individuals with IP measurement • The increase in power from improved phenotyping (better phenotypic homogeneity) could be less than decrease in power due to a smaller sample • Possibility of inconsistent recording of phenotypes across collaborating centers • Some of the proposed IP in psychiatry may not be less complex than the diagnoses • Danger of confusing biomarkers and epihenomena with IP (which are supposed to be involved in aetiology) • It is not clear which phenotypes could be valid and useful IP of psychiatric diseases.
246 Chapter 8
Type-II diabetes is an example of how an application of quantitative IP (in this case glycemic traits or BMI) complemented the typical case-control approach and led to new discoveries (e.g. association of GCK loci with fasting glucose levels) 80. To date, attempts to use suspected IP in the psychiatric genetics (either cross-diagnostically or within particular diagnoses) were less fruitful 81, although there were some encouraging examples. For instance, a GWAS of ASD, which, besides the diagnoses, used quantitative measures of autistic symptoms, found multiple previously unknown associations 82. In the field of SCZ, one study used a combined score of multiple variants associated or nominally associated with SCZ (polygenic risk score) and found a modest association with quantitative measures of psychosis 83, whereas another studies used the same score and detected association with total brain and white matter volume 84 or with working memory-related prefrontal brain activation 85. There are also examples of cross-diagnostic applications - Ruderfer et al. found an association of a BD polygenic risk score with dimension of mania in patients with SCZ 86. Investigation of additional phenotypes can be useful both at the genome-wide level (in the discovery phase) and in the follow-up studies of the associated variants. Bulik et al. reviewed and evaluated subphenotypes and potential IP in AN and EDs 74 (e.g. perfectionism, cognitive set-shifting, obssessionality or impulsivity), but the current state of research does not give a clear indication about which of them is the most promising. There are recent CGSs 87,88 and a GWAS 43,89 which investigated ED- related temperamental and psychometric phenotypes. None of them, however, found significant associations. To date, the search for relevant IP in EDs focused on behavioral and psychometric traits. In the future, all levels of human organism functioning need to be taken into account, including the transcriptome, proteome, neuroanatomy and neurological functioning (imaging genetics). It will become easier to determine valid IP once the knowledge about the aetiological pathways and mechanisms of AN improves. Studies of
247 Chapter 8 physiological, neurobiological or neurocognitive IP are likely to follow successful GWAS focused on diagnostic categories.
Conclusions
The early stage of GWASs in psychiatric disorders was considered largely unsuccessful. Lack of significant results came as a disappointment and criticism was raised that investing resources in GWASs does not pay off 90. At that time, much less was known about the underlying genetic architecture of psychiatric diseases; nowadays it is clear that these studies were unlikely to yield significant findings, predominantly due to their insufficient sample sizes. As many as 5 GWASs in SCZ found no significant SNP associations 32. A study which was first to report genome-wide significant findings used the data of 8,000 cases with SCZ and twice as many controls 64. Looking back at the year 2011, the accrual of robust findings in SCZ in relation to the increasing sample was even greater than expected 32. One of the reasons why the recent GWASs are increasingly successful (besides the massive samples sizes) is the rigorous, uniform and transparent methodology (methodological homogeneity). The field of genetics in AN and other EDs is still in an early stage. It has the advantage of being able to learn from the more advanced fields and, thus, divert the resources for research in optimal directions. Cooperation of GCAN with the Psychiatric Genomics Consortium (PGC) creates opportunities for cross-disorder analyses and is likely to greatly increase the pace of genetic discoveries in AN. It is likely to shed light on AN's connections with other comorbid and genetically related disorders (especially that the genotyping platform which will be used in the future research - Illumina's PsychArray BeadChip - was designed for studies of psychiatric disorders). Before that comes to fruition, however, sample sizes of more than 25,000 (at minimum) individuals with AN are necessary, so that a barrier of at least several genome-wide significant SNP associations is breached. More distant
248 Chapter 8 goals, envisioned by PGC, include amassing of 100,000 cases per disorder of interest. Further follow-up studies of the variants identified in a GWAS should be based on statistically significant and replicated results (suggestive association in an underpowered study is not enough). At this early stage of genetic research in AN, investing effort in genome-wide genotyping of large number of individuals with AN will probably result in more insight into biology than more costly sequencing of smaller number of individuals 35. Substantial progress in the understanding of the genetic substrate of AN and its relations to other diseases is bound to come. Nevertheless, patience is advised and the hopes should not be inflated by unrealistic promises. Such progress will require a tremendous effort and it will not be quick. Valuable achievements rarely come easily. Continuous collection of DNA samples and unified phenotypic characterization of patients is a conditio sine qua non. The Psychiatric Genomics Consortium-Anorexia Nervosa Working Group (PGC_AN) is already in place, and it constitutes a framework which can expedite and organize this process. The most important and challenging aspect of this job, however, lies on the part of the patients and clinicians. We encourage them to contribute to this effort, as GWASs are an indispensable step of what has already started to change the face of the psychiatry.
Readers who would like to learn about how to contribute are advised to contact the PGC at http://www.med.unc.edu/pgc.
Acknowledgements
Part of this work was supported by funding from the Marie Curie Research Training Network INTACT (Individually tailored stepped care for women with eating disorders; reference number: MRTN-CT-2006-035988).
249 Chapter 8
References
1. Boraska V, Franklin C, Floyd J, Thornton L, Huckins L, Southam L, et al. A genome-wide association study of anorexia nervosa. Mol Psychiatry 2014;19:1085-1094. 2. Keel PK, Klump KL. Are eating disorders culture-bound syndromes? implications for conceptualizing their etiology. Psychol Bull 2003;129:747- 769. 3. Lilenfeld LR, Kaye WH, Greeno CG, Merikangas KR, Plotnicov K, Pollice C, et al. A controlled family study of anorexia nervosa and bulimia nervosa: Psychiatric disorders in first-degree relatives and effects of proband comorbidity. Arch Gen Psychiatry 1998;55:603-610. 4. Strober M, Freeman R, Lampert C, Diamond J, Kaye W. Controlled family study of anorexia nervosa and bulimia nervosa: Evidence of shared liability and transmission of partial syndromes. Am J Psychiatry 2000;157:393-401. 5. Klump KL, Suisman JL, Burt SA, McGue M, Iacono WG. Genetic and environmental influences on disordered eating: An adoption study. J Abnorm Psychol 2009;118:797. 6. Bulik CM, Sullivan PF, Tozzi F, Furberg H, Lichtenstein P, Pedersen NL. Prevalence, heritability, and prospective risk factors for anorexia nervosa. Arch Gen Psychiatry 2006;63:305-312. 7. Klump KL, Miller K, Keel P, McGue M, Iacono W. Genetic and environmental influences on anorexia nervosa syndromes in a population based twin sample. Psychol Med 2001;31:737-740. 8. van Asselt KM, Kok HS, van der Schouw YT, Peeters PH, Pearson PL, Grobbee DE. Role of genetic analyses in cardiology: Part II: Heritability estimation for gene searching in multifactorial diseases. Circulation 2006;113:1136-1139. 9. Arcelus J, Witcomb GL, Mitchell A. Prevalence of eating disorders amongst dancers: A systemic review and meta-analysis. European Eating Disorders Review 2013;22:92-101.
250 Chapter 8
10. Vineis P, Pearce NE. Genome-wide association studies may be misinterpreted: Genes versus heritability. Carcinogenesis 2011;32:1295- 1298. 11. Yang J, Lee SH, Goddard ME, Visscher PM. GCTA: A tool for genome-wide complex trait analysis. The American Journal of Human Genetics 2011;88:76- 82. 12. Steinhausen H, Jakobsen H, Helenius D, Munk-Jorgensen P, Strober M. A nationwide study of the family aggregation and risk factors in anorexia nervosa over three generations. Int J Eat Disord 2014;48:1-8. 13. Knowlton RG, Cohen-Haguenauer O, Van Cong N, Frezal J, Brown VA, Barker D, et al. A polymorphic DNA marker linked to cystic fibrosis is located on chromosome 7. Nature 1985;318:380-382. 14. Gusella JF, Wexler NS, Conneally PM, Naylor SL, Anderson MA, Tanzi RE, et al. A polymorphic DNA marker genetically linked to huntington's disease. Nature 1983;306:234-238. 15. Huckins LM, Boraska V, Franklin CS, Floyd JA, Southam L, Boraska V, et al. Using ancestry-informative markers to identify fine structure across 15 populations of european origin. European Journal of Human Genetics 2014;22:1190-1200. 16. Munafo M, Stothart G, Flint J. Bias in genetic association studies and impact factor. Mol Psychiatry 2009;14:119-120. 17. Siontis KC, Patsopoulos NA, Ioannidis JP. Replication of past candidate loci for common diseases and phenotypes in 100 genome-wide association studies. European Journal of Human Genetics 2010;18:832-837. 18. Collins AL, Kim Y, Sklar P, International Schizophrenia Consortium, O'Donovan MC, Sullivan PF. Hypothesis-driven candidate genes for schizophrenia compared to genome-wide association results. Psychol Med 2012;42:607-616. 19. Clarke TK, Weiss ARD, Berrettini WH. The genetics of anorexia nervosa. Clin Pharmacol Ther 2011;91:181-188.
251 Chapter 8
20. Rask-Andersen M, Olszewski PK, Levine AS, Schioth HB. Molecular mechanisms underlying anorexia nervosa: Focus on human gene association studies and systems controlling food intake. Brain Res Rev 2010;62:147-164. 21. Hinney A, Volckmar A. Genetics of eating disorders. Curr Psychiatry Rep 2013;15:1-9. 22. Slof-Op't Landt MC, Furth EF, Meulenbelt I, Bartels M, Hottenga JJ, Slagboom PE, et al. Association study of the estrogen receptor I gene (ESR1) in anorexia nervosa and eating disorders: No replication found. Int J Eat Disord 2013;47:211-214. 23. Zhang C, Chen J, Jia X, Yu S, Jiang W, Zhang R, et al. Estrogen receptor 1 gene rs2295193 polymorphism and anorexia nervosa: New data and meta- analysis. Asia Pacific Psychiatry 2013;5:331-335. 24. Brandys MK, Slof-Op't Landt MC, van Elburg AA, Ophoff R, Verduijn W, Meulenbelt I, et al. Anorexia nervosa and the Val158Met polymorphism of the COMT gene: Meta-analysis and new data. Psychiatr Genet 2012;22:130- 136. 25. Brandys MK, Kas MJ, van Elburg AA, Ophoff R, Slof-Op't Landt MC, Middeldorp CM, et al. The Val66Met polymorphism of the BDNF gene in anorexia nervosa: New data and a meta-analysis. The World Journal of Biological Psychiatry 2013;14:441-451. 26. Pinheiro AP, Bulik CM, Thornton LM, Sullivan PF, Root TL, Bloss CS, et al. Association study of 182 candidate genes in anorexia nervosa. Am J Med Genet B Neuropsychiatr Genet 2010;153B:1070-1080. 27. Slof-O't Landt M, Meulenbelt I, Bartels M, Suchiman E, Middeldorp C, Houwing-Duistermaat J, et al. Association study in eating disorders: TPH2 associates with anorexia nervosa and self-induced vomiting. Genes, Brain and Behavior 2011;10:236-243. 28. Calati R, De Ronchi D, Bellini M, Serretti A. The 5-HTTLPR polymorphism and eating disorders: A meta- analysis. Int J Eat Disord 2011;44:191-199. 29. Mercader JM, Saus E, Aguera Z, Bayes M, Boni C, Carreras A, et al. Association of NTRK3 and its interaction with NGF suggest an altered cross-
252 Chapter 8 regulation of the neurotrophin signaling pathway in eating disorders. Hum Mol Genet 2008;17:1234-1244. 30. Vink T, Hinney A, Van Elburg A, van Goozen SH, Sandkuijl L, Sinke R, et al. Association between an agouti-related protein gene polymorphism and anorexia nervosa. Mol Psychiatry 2001;6:325-328. 31. Klein RJ, Zeiss C, Chew EY, Tsai JY, Sackler RS, Haynes C, et al. Complement factor H polymorphism in age-related macular degeneration. Science 2005;308:385-389. 32. Kim Y, Zerwas S, Trace SE, Sullivan PF. Schizophrenia genetics: Where next? Schizophr Bull 2011;37:456-463. 33. Wray NR, Yang J, Hayes BJ, Price AL, Goddard ME, Visscher PM. Pitfalls of predicting complex traits from SNPs. Nature Reviews Genetics 2013;14:507- 515. 34. Bulik-Sullivan B, Loh P, Finucane H, Ripke S, Yang J, Patterson N, et al. LD score regression distinguishes confounding from polygenicity in genome- wide association studies. Nat Genet 2015;47:291-295. 35. Visscher P, Goddard M, Derks E, Wray N. Evidence-based psychiatric genetics, AKA the false dichotomy between common and rare variant hypotheses. Mol Psychiatry 2012;17:474-485. 36. Iyegbe C, Campbell D, Butler A, Ajnakina O, Sham P. The emerging molecular architecture of schizophrenia, polygenic risk scores and the clinical implications for GxE research. Soc Psychiatry Psychiatr Epidemiol 2014;49:169-182. 37. Nakabayashi K, Komaki G, Tajima A, Ando T, Ishikawa M, Nomoto J, et al. Identification of novel candidate loci for anorexia nervosa at 1q41 and 11q22 in japanese by a genome-wide association analysis with microsatellite markers. J Hum Genet 2009;54:531-537. 38. Rincon G, Tengvall K, Belanger J, Lagoutte L, Medrano J, Andre C, et al. Comparison of buccal and blood-derived canine DNA, either native or whole genome amplified, for array-based genome-wide association studies. BMC Research Notes 2011;4:226.
253 Chapter 8
39. Wang K, Zhang H, Bloss CS, Duvvuri V, Kaye W, Schork NJ, et al. A genome-wide association study on common SNPs and rare CNVs in anorexia nervosa. Mol Psychiatry 2011;16:949-959. 40. de Kovel CGF, Trucks H, Helbig I, Mefford HC, Baker C, Leu C, et al. Recurrent microdeletions at 15q11.2 and 16p13.11 predispose to idiopathic generalized epilepsies. Brain 2010;133:23-32. 41. Traylor M, Farrall M, Holliday EG, Sudlow C, Hopewell JC, Cheng Y, et al. Genetic risk factors for ischaemic stroke and its subtypes (the METASTROKE collaboration): A meta-analysis of genome-wide association studies. The Lancet Neurology 2012;11:951-962. 42. Button KS, Ioannidis JP, Mokrysz C, Nosek BA, Flint J, Robinson ES, et al. Power failure: Why small sample size undermines the reliability of neuroscience. Nature Reviews Neuroscience 2013;14:365-376. 43. Boraska V, Davis OSP, Cherkas LF, Helder SG, Harris J, Krug I, et al. Genome-wide association analysis of eating disorder-related symptoms, behaviors, and personality traits. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics 2012;159B:803-811. 44. Fall T, Ingelsson E. Genome-wide association studies of obesity and metabolic syndrome. Mol Cell Endocrinol 2014;382:740-757. 45. Guo Y, Lanktree MB, Taylor KC, Hakonarson H, Lange LA, Keating BJ, et al. Gene-centric meta-analyses of 108 912 individuals confirm known body mass index loci and reveal three novel signals. Hum Mol Genet 2013;22:184-201. 46. Sullivan P. Don't give up on GWAS. Mol Psychiatry 2011;17:2-3. 47. Kabi A, Nickerson KP, Homer CR, McDonald C. Digesting the genetics of inflammatory bowel disease: Insights from studies of autophagy risk genes. Inflamm Bowel Dis 2012;18:782-792. 48. Ripke S, O'Dushlaine C, Chambert K, Moran JL, Kahler AK, Akterin S, et al. Genome-wide association analysis identifies 13 new risk loci for schizophrenia. Nat Genet 2013;45:1150-1159. 49. Schizophrenia Working Group of the Psychiatric Genomics Consortium. Biological insights from 108 schizophrenia-associated genetic loci. Nature 2014;511:421-427.
254 Chapter 8
50. Levinson DF, Mostafavi S, Milaneschi Y, Rivera M, Ripke S, Wray NR, et al. Genetic studies of major depressive disorder: Why are there no genome- wide association study findings and what can we do about it? Biol Psychiatry 2014;76:510-512. 51. Stahl EA, Wegmann D, Trynka G, Gutierrez-Achury J, Do R, Voight BF, et al. Bayesian inference analyses of the polygenic architecture of rheumatoid arthritis. Nat Genet 2012;44:483-489. 52. Guella I, Sequeira A, Rollins B, Morgan L, Myers RM, Watson SJ, et al. Evidence of allelic imbalance in the schizophrenia susceptibility gene ZNF804A in human dorsolateral prefrontal cortex. Schizophr Res 2014;152:111-116. 53. Giusti-Rodriguez P, Sullivan PF. The genomics of schizophrenia: Update and implications. J Clin Invest 2013;123:4557-4563. 54. Bulik-Sullivan B, Finucane HK, Anttila V, Gusev A, Day FR, Genomics Consortium P, et al. An atlas of genetic correlations across human diseases and traits. bioRxiv 2015;014498:doi:10.1101/014498. 55. Hudson JI, Hiripi E, Pope Jr. HG, Kessler RC. The prevalence and correlates of eating disorders in the national comorbidity survey replication. Biological Psychiatry 2007;61:348-358. 56. Keski-Rahkonen A, Hoek HW, Susser ES, Linna MS, Sihvola E, Raevuori A, et al. Epidemiology and course of anorexia nervosa in the community. Am J Psychiatry 2007;164:1259-1265. 57. Saha S, Chant D, Welham J, McGrath J. A systematic review of the prevalence of schizophrenia. Plos Medicine 2005;2:0413-0433. 58. Goldner EM, Hsu L, Waraich P, Somers JM. Prevalence and incidence studies of schizophrenic disorders: A systematic review of the literature. Can J Psychiatry 2002;47:833-843. 59. Power RA, Kyaga S, Uher R, MacCabe JH, Langstrom N, Landen M, et al. Fecundity of patients with schizophrenia, autism, bipolar disorder, depression, anorexia nervosa, or substance abuse vs their unaffected siblings. JAMA psychiatry 2013;70:22-30.
255 Chapter 8
60. Arcelus J, Mitchell AJ, Wales J, Nielsen S. Mortality rates in patients with anorexia nervosa and other eating disorders: A meta-analysis of 36 studies. Arch Gen Psychiatry 2011;68:724-731. 61. Saha S, Chant D, McGrath J. A systematic review of mortality in schizophrenia: Is the differential mortality gap worsening over time? Arch Gen Psychiatry 2007;64:1123-1131. 62. Uher R. The role of genetic variation in the causation of mental illness: An evolution-informed framework. Mol Psychiatry 2009;14:1072-1082. 63. Klump KL, Burt SA, McGue M, Iacono WG. Changes in genetic and environmental influences on disordered eating across adolescence: A longitudinal twin study. Arch Gen Psychiatry 2007;64:1409-1415. 64. International Schizophrenia Consortium, Purcell SM, Wray NR, Stone JL, Visscher PM, O'Donovan MC, et al. Common polygenic variation contributes to risk of schizophrenia and bipolar disorder. Nature 2009;460:748-752. 65. Fromer M, Pocklington AJ, Kavanagh DH, Williams HJ, Dwyer S, Gormley P, et al. De novo mutations in schizophrenia implicate synaptic networks. Nature 2014;506:179-184. 66. Malaspina D, Harlap S, Fennig S, Heiman D, Nahon D, Feldman D, et al. Advancing paternal age and the risk of schizophrenia. Arch Gen Psychiatry 2001;58:361-367. 67. Crow JF. Development. there's something curious about paternal-age effects. Science 2003;301:606-607. 68. Racine S, Culbert K, Burt S, Klump K. Advanced paternal age at birth: Phenotypic and etiologic associations with eating pathology in offspring. Psychol Med 2014;44:1029-1041. 69. Hoek HW, van Hoeken D. Review of the prevalence and incidence of eating disorders. International Journal of Eating Disorders 2003;34:383-396. 70. Purcell SM, Moran JL, Fromer M, Ruderfer D, Solovieff N, Roussos P, et al. A polygenic burden of rare disruptive mutations in schizophrenia. Nature 2014;506:185-190.
256 Chapter 8
71. Smoller JW, Craddock N, Kendler K, Lee PH, Neale BM, Nurnberger JI, et al. Identification of risk loci with shared effects on five major psychiatric disorders: A genome-wide analysis. Lancet 2013;381:1371-1379. 72. Andreassen O, Harbo H, Wang Y, Thompson W, Schork A, Mattingsdal M, et al. Genetic pleiotropy between multiple sclerosis and schizophrenia but not bipolar disorder: Differential involvement of immune-related gene loci. Mol Psychiatry 2014;20:207-214. 73. Cross-Disorder Group of the Psychiatric Genomics Consortium. Genetic relationship between five psychiatric disorders estimated from genome-wide SNPs. Nat Genet 2013;45:984-994. 74. Bulik CM, Hebebrand J, Keski-Rahkonen A, Klump KL, Reichborn- Kjennerud T, Mazzeo SE, et al. Genetic epidemiology, endophenotypes, and eating disorder classification. Int J Eat Disord 2007;40:S52-S60. 75. Eddy KT, Dorer DJ, Franko DL, Tahilani K, Thompson-Brenner H, Herzog DB. Diagnostic crossover in anorexia nervosa and bulimia nervosa: Implications for DSM-V. Am J Psychiatry 2008;165:245-250. 76. Grave RD. Eating disorders: Progress and challenges. Eur J Intern Med 2011;22:153-160. 77. Blinder BJ, Cumella EJ, Sanathara VA. Psychiatric comorbidities of female inpatients with eating disorders. Psychosom Med 2006;68:454-462. 78. Craddock N, Owen MJ. The kraepelinian dichotomy - going, going... but still not gone. Br J Psychiatry 2010;196:92-95. 79. Gottesman II, Gould TD. The endophenotype concept in psychiatry: Etymology and strategic intentions. Am J Psychiatry 2003;160:636-645. 80. Stolerman ES, Florez JC. Genomics of type 2 diabetes mellitus: Implications for the clinician. Nature Reviews Endocrinology 2009;5:429-436. 81. Glahn DC, Knowles EE, McKay DR, Sprooten E, Raventos H, Blangero J, et al. Arguments for the sake of endophenotypes: Examining common misconceptions about the use of endophenotypes in psychiatric genetics. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics 2014;165B:122-130.
257 Chapter 8
82. Hu VW, Addington A, Hyman A. Novel autism subtype-dependent genetic variants are revealed by quantitative trait and subphenotype association analyses of published GWAS data. PLoS One 2011;6:e19067. 83. Derks EM, Vorstman JA, Ripke S, Kahn RS, Ophoff RA, Schizophrenia Psychiatric Genomic Consortium. Investigation of the genetic association between quantitative measures of psychosis and schizophrenia: A polygenic risk score analysis. PloS one 2012;7:e37852. 84. Terwisscha van Scheltinga, Afke F, Bakker SC, van Haren NE, Derks EM, Buizer-Voskamp JE, Boos H, et al. Genetic schizophrenia risk variants jointly modulate total brain and white matter volume. Biol Psychiatry 2013;73:525- 531. 85. Kauppi K, Westlye LT, Tesli M, Bettella F, Brandt CL, Mattingsdal M, et al. Polygenic risk for schizophrenia associated with working memory-related prefrontal brain activation in patients with schizophrenia and healthy controls. Schizophr Bull 2014;41:736-746. 86. Ruderfer D, Fanous A, Ripke S, McQuillin A, Amdur R, Gejman P, et al. Polygenic dissection of diagnosis and clinical dimensions of bipolar disorder and schizophrenia. Mol Psychiatry 2013;19:1017-1024. 87. Jonassaint CR, Szatkiewicz JP, Bulik CM, Thornton LM, Bloss C, Berrettini WH, et al. Absence of association between specific common variants of the obesity-related FTO gene and psychological and behavioral eating disorder phenotypes. Am J Med Genet B Neuropsychiatr Genet 2011;156B:454-461. 88. Root TL, Szatkiewicz JP, Jonassaint CR, Thornton LM, Pinheiro AP, Strober M, et al. Association of candidate genes with phenotypic traits relevant to anorexia nervosa. Eur Eat Disord Rev 2011;19:487-493. 89. Wade TD, Gordon S, Medland S, Bulik CM, Heath AC, Montgomery GW, et al. Genetic variants associated with disordered eating. Int J Eat Disord 2013;46:594-608. 90. Couzin-Frankel J. Major heart disease genes prove elusive. Science 2010;328:1220-1221.
258 Chapter 8
91. Gauderman W, Morrison J. QUANTO 1.1: A computer program for power and sample size calculations for genetic-epidemiology studies. http://Biostats.usc.edu/quanto.html. 2006.
259 Chapter 8
Concluding remarks
The general aim of the research presented in this thesis was to improve understanding of AN and related conditions, such as other eating disorders and obesity. As with possibly all of the research in the biomedical sciences, it was performed with hope that its results would eventually add to the building of the way leading to improved prevention and treatments. Such general aims, in order to be pursued, must be reduced to particular research challenges. The studies included in this thesis focused specifically on searching for: -the genetic variants that increase or decrease the risk of developing AN (primary goal), -the genetic variants that play a role in the body-weight related parameters in the general population and the variants that might shed light on the hypothetical common genetic background of AN and obesity (secondary goal), -the biochemical markers of the disease and the genetic variants related to those biomarkers (secondary goal).
Initially, the investigated genes and their genetic variants were chosen on the basis of hypothesized aetiological pathways. Soon, however, our take shifted towards genome-wide approaches, rather than focusing on selected genes. The shift was necessary, as reflected by the results of the two meta-analyses of candidate-gene studies (concerning SNPs on BDNF and COMT genes) included in this thesis. These meta-analyses summarized evidence from many studies which looked into the same SNPs in AN and they also incorporated our own, novel data. Even though some of the studies included in the meta-analyses reported positive findings, the overall result showed no evidence of association between the selected SNPs and AN. One of the meanings of these disillusioning results was that a gene-association study focused on hand-picked polymorphisms is unlikely to succeed. That remains true, even though the selection of SNPs and genes had usually been
260 Chapter 8 based on compelling biological hypotheses and was inspired by other types of studies (e.g. studies of serum BDNF in AN, chapter 5). These results were in line with findings from several other psychiatric disorders and they underlined the conclusion that the genome-wide approach was inevitable to move the field forward. Chapter 2 presents a study which utilized the results of a genome- wide association studies (GWAS) from a different field - that of BMI and obesity 1,2. The best, high-confidence hits from these GWAS were tested for association with AN (in cases with AN and population controls). These SNPs turned out not to be associated in either a single-SNP analysis or in an analysis of a polygenic risk score (a combined effect of the SNPs in question). The difficulty with the GWA studies lies in the fact that they require a lot of resources and effort (not only money-wise, but also in terms of time and organization). This can be an important limiting factor. Even the largest GWAS in AN up to date, an unprecedented world-wide cooperation of many universities and other research centers (appendix) should have had a larger sample size. It appears though that at the time when the study was carried out, there were not many more DNA samples of patients with AN around the world. In other words, most of the world's DNA samples had already been included in the study (this statement is based on a personal observation). Furthermore, some of the DNA samples were of unsatisfactory quality and had to be removed from the analysis (which is a commonplace in this type of studies). Another obstacle which the researchers faced was the issue of the control sample. Due to the funding limitations the control samples were in sillico, which means that they were not genotyped along with the cases, but taken from already existing databases (except for two small control groups). Genome-wide data in combination with the modern statistics offer ways to analyze how well the case and control samples are matched, in order to ensure reliability of the analysis (mismatched samples produce unreliable results, due to e.g. population stratification or batch biases). As much as it could have been accounted for in the main case-control analysis, the lack of
261 Chapter 8
"original" controls turned out to be a major hindrance in the subsequent CNV analyses. Although this GWAS progressed the field of the genetic research in AN substantially, it appears that much greater scale of research is necessary before the field can begin moving on from the genetic association studies to functional investigation of identified variants and pathways. The variety of research questions that can be addressed with the genome-wide data extend far beyond the typical case-control association. Possible avenues for research include the analyses of structural variation. The manuscript in chapter 7 presents the analysis of selected large and rare CNVs in AN. Ideally, the case and the control DNA samples would be genotyped on the same genotyping platform, under the same conditions, and with random allocation of cases and controls across the batches. Unfortunately, most of the controls in the GWA study (appendix) came in sillico, they were genotyped on various platforms and there were no intensity data available (only the genotype calls). We managed to obtain the control data from additional databases, but these came in a form of either raw CNV calls (pre-QC) or QC-ed CNV calls and they were also genotyped with various genotyping Illumina chips. The analysis focused on predefined CNV regions and took several measures to ensure that the identified CNVs were reliable, but in light of the difficulties with the control samples we have decided not to send the manuscript for publication and consider its results as preliminary. Acknowledging all the caveats, the study found no association of the selected large and rare CNVs and AN. Interestingly, a study in progress which investigates CNVs in the same dataset of cases with AN and in a different control sample led to similar results3. The authors looked for CNVs larger than 100kb in the predefined "psychiatric" CNV regions (7 of which were the same as in our study) and found no signs of association with AN. The fact that it is so difficult to unravel the mechanisms underlying AN should not be surprising or discouraging. There are positive examples from related fields (such as the genetic research in schizophrenia, briefly discussed in chapter 8), which show that the amount of relevant knowledge
262 Chapter 8 gained from the gene-association studies increases along with the increase of the sample size (assuming that the other aspects reflecting the studies quality remain state-of-the-art). Even though the practical applications of the current findings in the genetics of AN are probably decades away, the scientific efforts should be pursued, since the prize is well worth the game.
References
1. Willer CJ, Speliotes EK, Loos RJ, Li S, Lindgren CM, Heid IM, et al. Six new loci associated with body mass index highlight a neuronal influence on body weight regulation. Nat Genet 2009;41:25-34. 2. Thorleifsson G, Walters GB, Gudbjartsson DF, Steinthorsdottir V, Sulem P, Helgadottir A, et al. Genome-wide association yields new sequence variants at seven loci that associate with measures of obesity. Nat Genet 2009;41:18- 24. 3. Yilmaz et al., 2015, abstract at the World Congress of Psychiatric Genetics 2015, work in progress.
263 Addendum
Addendum
English summary
Eating disorders (EDs) encompass anorexia nervosa (AN), bulimia nervosa, binge eating disorder, other specified feeding or eating disorders, unspecified feeding or eating disorders and, although it is a matter of a debate, obesity. These disorders entail a substantial socioeconomic burden 1. AN, which is the main focus of this thesis, is notorious for the highest standardized mortality ratio in all psychiatric illnesses (its mortality rate is 6 to 10 times higher than that in the general population 2,3. It has been established in multiple studies that a heritable component plays an important role in the aetiology of AN. Twin and adoption studies determined that genetic factors are responsible for 46 to 78 percent of variance in AN 4-6. A 10-fold increase in lifetime risk of AN in a first-degree female relative of a person with an ED (comparing to relatives of unaffected individuals) was shown in a family study 7. Notwithstanding the results of those formal genetic studies, the particular genetic factors associated with the aetiology and maintenance of AN and other EDs remain unknown. The reasons for searching for those factors are manifold. The ultimate (and very remote) goals are to cure or prevent the disorder, alleviate individual's suffering, improve their quality of life and reduce the socioeconomic burden. Before these can be pursued, the genetic variants which are associated with AN need to be pin-pointed, which in turn leads to follow-up studies of those variants or biological pathways in which they are involved. Understanding of the mechanisms of the disease might eventually allow development of better (tailored) psycho- and pharmacotherapies. The research in this thesis focused on searching for: -the genetic variants which increase or decrease the risk of developing AN (primary goal) -the variants which play a role in the body-weight related parameters in the general population
264 Addendum
-the variants which might shed light on the hypothetical common genetic background of AN and obesity -the biochemical markers of the disease and the genetic variants related to them.
In general, the methods used in the presented thesis include tests for association between genetic variants and AN (a classical case-control scenario, where statistical tests of difference in frequency of a genetic variant in cases vs. controls determine association) and tests of association with a continuous phenotype (such as e.g. BMI). In chapters 4, 5 and 6 we also use a meta-analytic framework to aggregate and assess the data from other studies and (in chapter 4 and 6) combine it with our own, novel data.
Single-nucleotide polymorphisms (SNPs) associated with body-mass index (BMI) in the general population were tested for association with AN, both on an individual basis and as a combined score (a so-called polygenic risk score). There was no evidence of association with AN in both scenarios. This is an indication that, on the level of genetic aetiology, AN is not on the same continuum as BMI in the general population. We also studied genetic variants in the POMC gene locus (part of the melanocortin system) and its association to several body-weight and food intake related phenotypes. Signals of association with waist:hip ratio, visceral fat and abdominal fat made POMC a strong candidate for further studies. Nevertheless, these results should be taken with caution in light of a small sample size. We further focused on two SNPs which are well-known in the genetic research in psychiatry. These SNPs are located on the BDNF and COMT genes, which both have theoretical and empirical support for playing a role in the aetiology of psychiatric disorders. Each of those SNPs was tested several times in previous studies in AN, but the results were inconsistent. We applied the meta-analytic framework to evaluate and combine the evidence from those previous studies with our own, novel data. This work showed that
265 Addendum
SNPs rs6265 (BDNF) and rs4680 (COMT) were not associated with AN. Our results resolve the controversy stemming from prior inconsistent results and underlie the necessity for large sample sizes and stringent quality control and analysis of the data. Brain-derived neurotrophic factor (BDNF) belongs to a family of neurotrophins and has multiple functions which might theoretically play a role in the development and maintenance of AN. Here, we combined the data from several studies and performed a meta-analysis on the levels of BDNF protein in the blood serum of patients with AN. This study showed that the serum BDNF level in patients with AN is lower than in healthy controls. Copy-number variants (CNVs) are a form of genetic variation which encompasses more than a single nucleotide. Several large and recurrent CNVs have been associated with psychiatric disorders (such as intellectual disability, schizophrenia or epilepsy) or obesity. We tested preselected CNVs in cases with AN versus controls, using the genome-wide data from the AN GWAS study 8. None of the tested CNVs appeared to be associated with AN, although there were important caveats which needed to be taken into account when interpreting these results. A genome-wide association study of SNPs in AN and controls was part of this thesis work. Even though it was the largest study of its kind in AN, no variants associated at a genome-wide significance level were found.
The current accomplishments of the gene-association research in AN are more modest, compared to the research in some other psychiatric disorders, such as schizophrenia. The main reason for that are the smaller numbers of the available DNA samples from patients with AN. The genetic architecture of AN suggest that (similarly as in schizophrenia) when the sample sizes increase substantially, new loci harbouring variants associated with susceptibility to AN will be determined.
References 1. Simon J, Schmidt U, Pilling S. The health service use and cost of eating disorders. Psychol Med 2005;35:1543-1551.
266 Addendum
2. Birmingham CL, Su J, Hlynsky JA, Goldner EM, Gao M. The mortality rate from anorexia nervosa. Int J Eat Disord 2005;38:143-146. 3. Papadopoulos FC, Ekbom A, Brandt L, Ekselius L. Excess mortality, causes of death and prognostic factors in anorexia nervosa. The British Journal of Psychiatry 2009;194:10-17. 4. Thornton LM, Mazzeo SE, Bulik CM. The heritability of eating disorders: Methods and current findings. In: Behavioral neurobiology of eating disorders. Springer, 2011, p. 141-156. 5. Kortegaard LS, Hoerder K, Joergensen J, Gillberg C, Kyvik KO. A preliminary population- based twin study of self-reported eating disorder. Psychol Med 2001;31:361-365. 6. Wade TD, Bulik CM, Neale M, Kendler KS. Anorexia nervosa and major depression: Shared genetic and environmental risk factors. Am J Psychiatry 2000;157:469-471. 7. Strober M, Freeman R, Lampert C, Diamond J, Kaye W. Controlled family study of anorexia nervosa and bulimia nervosa: Evidence of shared liability and transmission of partial syndromes. Am J Psychiatry 2000;157:393-401. 8. Boraska V, Franklin C, Floyd J, Thornton L, Huckins L, Southam L, et al. A genome-wide association study of anorexia nervosa. Mol Psychiatry 2014;19:1085-1094.
267 Addendum
Nederlandse samenvatting
Eetstoornissen omvatten anorexia nervosa (AN), boulimia nervosa, de eetbuistoornis, andere gespecificeerde of ongespecificeerde voedings- en eetstoornissen en, ook al is het een kwestie van debat, obesitas. Deze aandoeningen leiden tot een aanzienlijke sociaal-economische last 1. AN, de belangrijkste focus van dit proefschrift, is berucht vanwege het hoogste gestandaardiseerde sterftecijfer in verhouding met alle psychiatrische aandoeningen (het sterftecijfer is 6 tot 10 keer hoger dan dat van de algemene bevolking 2, 3). Meerdere onderzoeken laten een belangrijke rol zien van erfelijke factoren in de etiologie van AN. Tweeling en adoptie studies tonen aan dat genetische factoren verantwoordelijk zijn voor 46 tot 78 procent van de variantie in AN 4-6. Er is een 10-voudige toename in het risico van het krijgen van een eetstoornis in een eerste graad vrouwelijk familielid van een persoon met een eetstoornis (in vergelijking met familieleden van niet aangetaste individuen) 7. Ondanks de resultaten van genetische studies zijn de specifieke genetische factoren die samenhangen met de etiologie, het ontstaan en het voortduren van AN en andere eetstoornissen nog onbekend. De redenen voor het zoeken naar die genetische factoren zijn legio. Het ultieme doel is het genezen of voorkomen van de ziekte, naast het lijden van het individu verlichten, het verbeteren van de kwaliteit van leven en vermindering van de sociaal-economische last. Voordat deze doelen kunnen worden gehaald, is het essentieel de genetische variaties die gekoppeld zijn AN te ontrafelen; hetgeen zal leiden tot vervolgonderzoek naar die variaties en de biologische routes waarbij zij betrokken zijn. Begrip van de mechanismen van de ziekte kan uiteindelijk de ontwikkeling van betere (op maat) psycho- en farmacotherapie mogelijk maken. Het onderzoek in dit proefschrift richtte zich op het zoeken naar: -de genetische variaties die het risico verhogen op het ontwikkelen van AN (primaire doel) -de variaties die een rol spelen bij lichaamsgewicht-gerelateerde factoren bij de algemene bevolking -de variaties die licht zouden kunnen werpen op de hypothetische gemeenschappelijke genetische achtergrond van AN en obesitas
268 Addendum
-de biochemische markers en de genetische varianten die betrekking hebben op AN.
In het algemeen omvatten de in dit proefschrift gebruikte methodes associatie testen tussen genetische varianten en AN (een klassiek case- control scenario, waarbij verschil in frequentie van een genetische variant in AN patiënten vs. controles statistische getest wordt) en tests voor het verband met een continue variabele (zoals bijvoorbeeld BMI) binnen een populatie. In de hoofdstukken 4, 5 en 6 wordt een meta-analyse beschreven met gegevens uit andere studies en (in hoofdstuk 4 en 6) gecombineerd met eigen verkregen nieuwe data.
Single nucleotide polymorphisms (SNPs) geassocieerd met body- mass index (BMI) in de algemene populatie werden getest op associatie met AN, zowel op individuele basis als een gecombineerde score (een zogeheten polygeen risico score). Er werd geen bewijs gevonden voor associatie met AN in beide scenario's. Dit is een indicatie dat de genetische etiologie in AN niet op hetzelfde continuüm ligt als BMI in de algemene bevolking. We bestudeerden ook genetische variaties in het POMC gen locus (onderdeel van het melanocortinesysteem) en de associatie met een aantal lichaamsgewicht en voedselinname verwante fenotypes. Aanwijzingen voor associatie met de taille: heup ratio, visceraal vet en buikvet maakten POMC een sterke kandidaat voor verdere studies. Toch manen deze resultaten tot voorzichtigheid over de relatie van POMC met deze parameters omdat het een kleine steekproef betrof. We richtten ons verder op twee SNPs die in het genetisch onderzoek in de psychiatrie bekend zijn. Deze SNPs zijn gelegen op het BDNF en het COMT gen. Voor beiden is theoretische en empirische onderbouwing voor een rol in de etiologie van psychiatrische stoornissen. Elk van deze SNPs werd meerdere malen in eerdere studies getest voor associatie met AN, maar de resultaten waren inconsistent. We voerden een meta-analyse uit en combineerden het bewijs van eerdere studies met onze eigen, nieuwe data. Uit dit werk bleek dat SNPs rs6265 (BDNF) en rs4680 (COMT) niet geassocieerd konden worden met AN. Onze resultaten lossen daarmee een van de controversen op die voortvloeien uit eerdere inconsistente
269 Addendum resultaten. We onderstrepen hiermee de noodzaak voor grote steekproeven en strenge kwaliteitscontrole en analyse van dit soort bevindingen. BDNF (BDNF) behoort tot een familie van neurotrofinen. Diverse aspecten van de functies van BDNF zouden theoretisch een rol kunnen spelen in de ontwikkeling en het onderhoud van AN. We combineerden de gegevens uit verschillende studies tot een meta-analyse van de niveaus van BDNF eiwit in het bloed (serum) van patiënten met AN. Deze studie toonde aan dat het serum BDNF niveau van patiënten met AN lager is dan bij gezonde controles. Copy-number variaties (CNV) zijn een vorm van genetische variatie die meer dan één nucleotide omvatten, waarbij stukken DNA ontbreken of gedupliceerd zijn. Verscheidene grote en terugkerende CNVs zijn geassocieerd met psychiatrische stoornissen (zoals intellectuele beperking, schizofrenie of epilepsie) of obesitas. We testten voorgeselecteerde CNVs in patiënten met AN versus controles, met behulp van genoom-brede data van de AN GWAS studie 8. Geen van de geteste CNVs leek geassocieerd. Wel moeten er belangrijke kanttekeningen worden geplaatst bij het interpreteren deze resultaten. Een genoom-brede associatie studie van SNPs in AN en controles maakte deel uit van dit proefschrift. Ook al was het de grootste studie in zijn soort in AN, geen van de variaties bleken geassocieerd met een genoom- breed significantie niveau. De huidige resultaten van het gen-associatie onderzoek bij AN zijn bescheiden in vergelijking met het onderzoek naar andere psychiatrische stoornissen, zoals schizofrenie. De belangrijkste reden dat er bij genetisch onderzoek naar AN minder resultaat is geboekt is te wijten aan de kleinere aantallen beschikbare DNA-monsters van patiënten met AN. Als het aantal DNA samples in genetische studies naar AN aanzienlijk wordt verhoogd (zoals bij schizofrenie), is de verwachting dat nieuwe loci worden gevonden die variaties herbergen die geassocieerd zijn met gevoeligheid voor AN.
References 1. Simon J, Schmidt U, Pilling S. The health service use and cost of eating disorders. Psychol Med 2005;35:1543-1551.
270 Addendum
2. Birmingham CL, Su J, Hlynsky JA, Goldner EM, Gao M. The mortality rate from anorexia nervosa. Int J Eat Disord 2005;38:143-146. 3. Papadopoulos FC, Ekbom A, Brandt L, Ekselius L. Excess mortality, causes of death and prognostic factors in anorexia nervosa. The British Journal of Psychiatry 2009;194:10-17. 4. Thornton LM, Mazzeo SE, Bulik CM. The heritability of eating disorders: Methods and current findings. In: Behavioral neurobiology of eating disorders. Springer, 2011, p. 141-156. 5. Kortegaard LS, Hoerder K, Joergensen J, Gillberg C, Kyvik KO. A preliminary population- based twin study of self-reported eating disorder. Psychol Med 2001;31:361-365. 6. Wade TD, Bulik CM, Neale M, Kendler KS. Anorexia nervosa and major depression: Shared genetic and environmental risk factors. Am J Psychiatry 2000;157:469-471. 7. Strober M, Freeman R, Lampert C, Diamond J, Kaye W. Controlled family study of anorexia nervosa and bulimia nervosa: Evidence of shared liability and transmission of partial syndromes. Am J Psychiatry 2000;157:393-401. 8. Boraska V, Franklin C, Floyd J, Thornton L, Huckins L, Southam L, et al. A genome-wide association study of anorexia nervosa. Mol Psychiatry 2014;19:1085-1094.
271 Addendum
Streszczenie w języku polskim
Na kategorię zaburzeń składają się: anorexia nervosa (AN), bulimia nervosa, jedzenie napadowe, inne specyficzne zaburzenia żywienia oraz niespecyficzne zaburzenia żywienia. Niektórzy badacze dodają również do tej grupy otyłość. Powyższe zaburzenia pociągają za sobą poważne koszty socjoekonomiczne 1. AN, która stanowi główny przedmiot zainteresowania niniejszej pracy doktorskiej, jest znana ze względu na najwyższy standaryzowany współczynnik umieralności wśród wszystkich zaburzeń psychiatrycznych (śmiertelność wyższa od 6 do 10 razy aniżeli w populacji ogólnej 2,3). Zostało potwierdzone w szeregu badań genetycznych, że czynnik odziedziczalny odgrywa istotną rolę w etiologii AN. W badaniach bliźniaków i adopcyjnych ustalono, że czynniki genetyczne odpowiadają za 46 do 78 procent zmienności w AN 4-6. Z kolei w innym badaniu rodzinnym wykazano, że będąca kobietą krewna pierwszego stopnia osoby z zaburzeniem żywienia ma dziesięciokrotnie zwiększone ryzyko wystąpienia AN w ciągu życia (w porównaniu do krewnych osób niedotkniętych zaburzeniami żywienia) 7. Pomimo jednoznacznych wyników formalnych badań genetycznych, konkretne warianty genetyczne związane z etiologią i przebiegiem AN oraz innych zaburzeń żywienia pozostają nieznane. Motywacja do poszukiwania owych wariantów jest wieloraka. Ostatecznym (i jednocześnie bardzo odległym) celem jest wyleczenie zaburzenia lub jego zapobieżenie, zmniejszenie cierpienia pacjenta i podniesienie jego jakości życia, a także ograniczenie kosztów socjoekonomicznych. Na drodze do tych odległych celów konieczne jest określenie wariantów genetycznych związanych z AN, co z kolei pozwoli na badania funkcjonalne tych wariantów oraz szlaków biologicznych, w które są one zaangażowane. Zrozumienie biologicznych podstaw choroby da szansę na opracowanie lepszych (indywidualnie dopasowanych) psycho- i farmakoterapii. Badania przedstawione w poniższej pracy doktorskiej skupiają się na poszukiwaniu:
272 Addendum
-wariantów genetycznych, które zwiększają lub zmniejszają ryzyko wystąpienia AN (cel główny) -wariantów, które mają związek z różnymi parametrami masy ciała w populacji ogólnej -wariantów, które mogą rzucić nowe światło na hipotetyczne wspólne tło genetyczne AN i otyłości -biochemiczne markery AN oraz związane z nimi warianty genetyczne
Ogólnie rzecz biorąc, metody zastosowane w niniejszej pracy doktorskiej to testy asocjacji między wariantami genetycznymi (klasyczne badanie typu grupa badawcza vs grupa kontrolna, gdzie istotna statystycznie różnica w częstotliwości występowanie wariantu między grupami świadczy o związku), oraz testy asocjacji z fenotypem ciągłym (jak np. body-mass index, BMI). W rozdziałach 4, 5 i 6 stosujemy również meta-analizę, po to aby zespolić i ocenić dane z innych badań oraz (w rozdziale 4 i 6) połączyć je z naszymi własnymi danymi.
Polimorfizmy pojedynczego nukleotydu (SNP) związane ze współczynnikiem BMI w populacji ogólnej zostały przetestowane pod względem asocjacji z AN, zarówno na poziomie każdego polimorfizmu, jak i w postaci powiązanej (tzw. polygenic risk score - poligeniczny współczynnik ryzyka). W obu przypadkach nie znaleziono dowodów na powiązanie z AN. Rezultaty te wskazują, że na poziomie etiologii genetycznej AN nie znajduje się na tym samym kontinuum co BMI w populacji ogólnej. Uwagę badawczą skupiliśmy również na genie POMC (część układu melanokortyn) oraz jego asocjacji z wybranymi fenotypami z zakresu masy ciała oraz żywienia. Sygnały asocjacji zostały wykryte pomiędzy SNPami z locus POMC oraz współczynnikiem talia:biodra, poziomem wisceralnej tkanki tłuszczowej oraz poziomem brzusznej tkanki tłuszczowej. Rezultaty te czynią gen POMC interesującym obiektem do dalszych badań, aczkolwiek należy traktować je z ostrożnością ze względu na niewielką próbę badawczą.
273 Addendum
W następnej kolejności przyjrzeliśmy się dwóm SNPom, które są dobrze znane w świecie psychiatrii genetycznej. Chodzi o SNPy ulokowane na genach BDNF oraz COMT. Istnieją empiryczne i teoretyczne przesłanki związku tych genów z zaburzeniami psychiatrycznymi. Oba SNPy były już wcześniej kilkukrotnie badane pod kątem związku z AN przez innych autorów, ale wyniki pozostały niejednoznaczne. Zastosowaliśmy podejście meta-analityczne aby zagregować dane z tych poprzednich badań z naszymi nowymi danymi. Praca ta wykazała, że SNPy rs6265 (BDNF) oraz rs4680 (COMT) nie są powiązane z AN. Rozstrzyga to kontrowersje wynikające z poprzednich badań i jednocześnie podkreśla konieczność stosowania dużych grup badawczych oraz najwyższych standardów kontroli jakości oraz analizy danych. Neurotropowy czynnik pochodzenia mózgowego (brain-derived neurotrophic factor, BDNF) należy do rodziny czynników wzrostu neuronów i pełni szereg funkcji, które teoretycznie mogą odgrywać rolę w etiologii AN. Żeby przyjrzeć się poziomowi BDNF w surowicy krwi pacjentek z AN przeprowadziliśmy meta-analizę danych z kilku badań tego właśnie parametru. Studium to wykazało, że poziom BDNF w surowicy krwi pacjentów z AN jest istotnie niższy, w porównaniu do próby kontrolnej. Genetyczne warianty strukturalne (warianty ilości kopii, copy- number variants, CNV) stanowią formę wariacji, która dotyczy więcej niż jednego nukleotydu. Stwierdzona została asocjacja kilku dużych i powtarzalnych CNV z chorobami psychiatrycznymi, takimi jak upośledzenie umysłowe, schizofrenia czy epilepsja, a także z otyłością. Przetestowaliśmy te wybrane CNV w populacji pacjentów z AN (vs próba kontrolna), korzystając z danych wygenerowanych w badaniu asocjacyjnym całego genomu 8. Żaden z rozpatrywanych wariantów nie okazał się związany z AN, chociaż przy interpretacji tych wyników pod uwagę wziąć trzeba kilka istotnych ograniczeń tego badania. Badanie asocjacyjne całego genomu wśród pacjentów z AN oraz populacji kontrolnej również stanowi część niniejszej dysertacji. Mimo tego, że było to największe tego rodzaju badanie w AN, nie wykryto żadnych
274 Addendum wariantów, które byłyby w sposób statystycznie istotny (na poziomie istotności właściwym dla badań całego genomu) związane z AN. Obecne osiągnięcia genetycznych badań asocjacyjnych w AN wyglądają skromniej od osiągnięć badań w niektórych innych zaburzeniach psychiatrycznych (w porównaniu z np. schizofrenią). Główną przyczyną takiego stanu rzeczy znacznie mniejsza liczba próbek DNA pacjentów z AN. Genetyczna architektura AN sugeruje, że (podobnie jak w schizofrenii) wraz ze wzrostem wielkości próby badawczej odkryte zostaną nowe warianty genetyczne związane z ryzykiem wystąpienia AN.
References 1. Simon J, Schmidt U, Pilling S. The health service use and cost of eating disorders. Psychol Med 2005;35:1543-1551. 2. Birmingham CL, Su J, Hlynsky JA, Goldner EM, Gao M. The mortality rate from anorexia nervosa. Int J Eat Disord 2005;38:143-146. 3. Papadopoulos FC, Ekbom A, Brandt L, Ekselius L. Excess mortality, causes of death and prognostic factors in anorexia nervosa. The British Journal of Psychiatry 2009;194:10-17. 4. Thornton LM, Mazzeo SE, Bulik CM. The heritability of eating disorders: Methods and current findings. In: Behavioral neurobiology of eating disorders. Springer, 2011, p. 141-156. 5. Kortegaard LS, Hoerder K, Joergensen J, Gillberg C, Kyvik KO. A preliminary population- based twin study of self-reported eating disorder. Psychol Med 2001;31:361-365. 6. Wade TD, Bulik CM, Neale M, Kendler KS. Anorexia nervosa and major depression: Shared genetic and environmental risk factors. Am J Psychiatry 2000;157:469-471. 7. Strober M, Freeman R, Lampert C, Diamond J, Kaye W. Controlled family study of anorexia nervosa and bulimia nervosa: Evidence of shared liability and transmission of partial syndromes. Am J Psychiatry 2000;157:393-401. 8. Boraska V, Franklin C, Floyd J, Thornton L, Huckins L, Southam L, et al. A genome-wide association study of anorexia nervosa. Mol Psychiatry 2014;19:1085-1094.
275 Addendum
Curriculum Vitae
Marek Kajetan Brandys was born on the 27th of November 1983 in Kraków, Poland. Across two millennia, he attended VLO high school in Kraków. In year 2002 Marek went on to study for MSc in applied psychology at the Jagiellonian University in Kraków. In the fifth and final year at the university he participated in the Socrates/Erasmus programme which allowed him to spend five fruitful months at the University of Utrecht. It was there that he met dr. Unna Danner, who was supervising a students' research project on the subject of symptoms of eating disorders (ED) in the general population. Prior to that, Marek had already taken interest in the intricacies of body image in the healthy and psychopathological populations, which was reflected in the topic of his master's thesis (body image and ED symptoms in people working out at gyms). Dr. Danner later got Marek in touch with prof. dr. Roger Adan. Prof. Adan was looking for someone to fill a position of a junior researcher and a PhD student in a research programme based on funding from Marie Curie Research Training Network grant. Marek got employed and as a result he spent three wonderful and productive years at UMC Utrecht, under the supervision of prof. Adan and prof. Annemarie van Elburg. After the funding expired, Marek was offered a year-long position by dr. Martien Kas, also at UMC Utrecht. At that time, he was also supervised by dr. Caroline de Kovel. All in all, it allowed him to continue his previous line of research and expand into new directions. Research performed over these four years in Utrecht led to the present PhD thesis. Currently Marek runs an own business in a touristic area close to his home city, Kraków.
276 Addendum
List of publications
Brandys MK, de Kovel CG, Kas MJ, van Elburg AA, Adan RA. Overview of genetic research in anorexia nervosa: The past, the present and the future. Int J Eat Disord 2015.
Bulik-Sullivan B, Finucane HK, Anttila V, Gusev A, Day FR, Genomics Consortium P, et al. An atlas of genetic correlations across human diseases and traits. bioRxiv 2015;014498:doi:10.1101/014498.
Kostrzewa E, Brandys M, van Lith H, Kas M. A candidate syntenic genetic locus is associated with voluntary exercise levels in mice and humans. Behav Brain Res 2015;276:8-16.
Boraska V, Franklin C, Floyd J, Thornton L, Huckins L, Southam L, et al. A genome-wide association study of anorexia nervosa. Mol Psychiatry 2014;19:1085-1094.
Huckins LM, Boraska V, Franklin CS, Floyd JA, Southam L, Boraska V, et al. Using ancestry-informative markers to identify fine structure across 15 populations of european origin. European journal of human genetics 2014;22:1190-1200. de Mooij-van Malsen J, van Lith H, Laarakker M, Brandys M, Oppelaar H, Collier D, et al. Cross-species genetics converge to TLL2 for mouse avoidance behavior and human bipolar disorder. Genes, Brain and Behavior 2013;12:653-657.
Brandys MK, Kas MJ, van Elburg AA, Ophoff R, Slof-Op't Landt MC, Middeldorp CM, et al. The Val66Met polymorphism of the BDNF gene in anorexia nervosa: New data and a meta-analysis. The World Journal of Biological Psychiatry 2013;14:441-451.
277 Addendum
Brandys MK, Slof-Op't Landt MC, van Elburg AA, Ophoff R, Verduijn W, Meulenbelt I, et al. Anorexia nervosa and the Val158Met polymorphism of the COMT gene: Meta-analysis and new data. Psychiatr Genet 2012;22:130- 136.
Brandys MK, Kas MJH, van Elburg AA, Campbell IC, Adan RAH. A meta- analysis of circulating BDNF concentrations in anorexia nervosa. World J Biol Psychiatry 2011.
Slof-Op 't Landt MCT, Meulenbelt I, Bartels M, Suchiman E, Middeldorp CM, Houwing-Duistermaat JJ, et al. Association study in eating disorders: TPH2 associates with anorexia nervosa and self-induced vomiting. Genes Brain Behav 2011;10:236-243.
Ternouth A, Brandys MK, van der Schouw YT, Hendriks J, Jansson J, Collier D, et al. Association study of POMC variants with body composition measures and nutrient choice. Eur J Pharmacol 2011.
Brandys MK, van Elburg AA, Loos RJF, Bauer F, Hendriks J, van der Schouw YT, et al. Are recently identified genetic variants regulating BMI in the general population associated with anorexia nervosa? American Journal of Medical Genetics Part B: Neuropsychiatric Genetics 2010;153B:695-699.
278 Addendum
Acknowledgements
When I moved to the Netherlands to pursue a scientific career, I was accompanied by the feelings of excitement and anxiety. It felt like stepping into the unknown. I was going to perform research in the biosciences, whereas my educational background and prior scientific endeavours were in the field of psychology. Living abroad for a longer term is always a deep experience, with consequences for one's personal development. My scientific and personal development was possible thanks to many great people that I had the privilege to meet.
First and foremost, I wish to express my gratefulness to prof. dr. Roger Adan, who was the primary supervisor of this PhD project. Dear Roger, thank you for seeing the potential and trusting in me at the beginning, for your scientific insight and knowledge, unceasing enthusiasm and many great discussions, which always led to new ideas and motivation boost. Prof. dr. Annemarie van Elburg, dr. Carolien de Kovel and dr. Martien Kas were supervisors who thought me how to "do science", created opportunities for it and were always there to help in times of need. When after three years the contract based on the Marie-Curie grant expired, dr. Kas decided to hire me for one more year. It allowed me to work on new projects in his group and, at the same time, make progress in terms of the PhD thesis. Thank you dr. Unna Danner for sharing your skills and expertise and for putting forward my candidacy for the PhD vacancy. Thanks to Nicole and all the members of the URGE group.
I would like to thank the reading committee consisting of prof. dr. Yvonne van der Schouw, prof. dr. Leon Kenemans, prof. dr. Wijbrand Hoek, prof. dr. Eric van Furth and prof. dr. Jan Veldink for having the patience to read this thesis and evaluate it critically.
279 Addendum
Thank you to the supervisors and members of the INTACT research- training network for creating a unique environment for learning, networking and enjoying life. In particular, I thank prof. Hans Kordy and dr. Stephanie Bauer for making INTACT possible. During 4 years in Utrecht I met many excellent scientists who directly or indirectly contributed to this thesis and I would like to thank them as well. These people include prof. Roel Ophoff, dr. Bobby Koeleman, prof. David Collier, prof. Eleftheria Zegginni, prof. Cindy Bulik, dr. Ross Crosby, dr. Filip Rybakowski, prof. Janet Treasure, prof. Eric Stice and many others. Thank you to all the officemates and people from the groups of Roger Adan and Martien Kas. In particular, I wish to say thanks to Ela, Jules, Judith, Sanna, Eneda, Anne, Ria, Myrte, Esther, Frank and Leo. Thank you for the great times, Asheeta, Sjoerd, Sietske, Marcin R., Vesna and Marcin B.. Another thank you goes to the administrative staff, especially Ria, Krista, Sandra and Vicki for having the patience. I am thankful to all the co-authors of the papers in which I was involved and also to the people affected by eating disorders who contributed their blood samples for research.
My parents, Wanda i Andrzej. Thank you for being the best parents in the world. I am grateful for your believing in me, for the support, for prioritizing education, for the work and hope, for your axiological systems. Thanks to my brother, Jacek, who is someone I can always depend on. Thanks Ewelina and Hania for bringing so much life into our home. I would also like to say thank you to the other members of my family and to my best friends from Poland. Thank you Kasia S. for fuelling my motivation. Thanks to Krzysiek, Tania, Marcin and Ewka for the good times we had together. I regret that many of you don't live in Poland these days but I know that is for the best.
I am also grateful to my country for giving me the education and to the European Union, which gave me the opportunity to participate in the
280 Addendum
Erasmus students' exchange program, which subsequently led to my joining the Marie-Curie network and becoming a PhD student in Utrecht.
281